Mission The Department s mission is communicated in three dimensions. It will endeavour

Transcription

1

2 Mission The Department s mission is communicated in three dimensions. It will endeavour to nurture graduates who will become leaders and professionals with a global outlook, ready to serve in the society of tomorrow with advanced knowledge and skills in computing and related areas; to conduct world-class research and promote inter-disciplinary collaboration, expanding the horizon of knowledge discovery and technology advancement; and to contribute and deliver professional services to the community at large with strong partnership and collaboration. Vision The Department is determined to assert its position as the leader in inter-disciplinary research and education in computing, generating worldwide impact and benefit to mankind.

3 Contents Message from Head Research Strengths Lists of Projects Highlights of Projects 02

4 Message from Head The Department of Computing is embracing a remarkable occasion in 2014 the 40 th anniversary. Over the past 40 years, the Department has made significant achievements in our grand missions of teaching, research and knowledge transfer to the profession and community. Our achievements have been recognized internationally as evident by being ranked among the top 50 in QS World University Subject Rankings in Computer Science and Information Systems in 2013 and We are now committed to becoming the leader in multi-disciplinary computing education and research and being the influential icon in smart computing. Our research vitality is reflected by our impressive records of highly competitive research grants, high quality publications in prestigious journals and top international conferences as well as strong industrial collaboration ties. With the aim of pushing the frontiers of computing and information technology, the Department has recently established a 3-tier framework with a common research infrastructure for cross-area and inter-disciplinary research. Tier 1 represents the departmentdriven, strategically focused research initiatives, currently in the two major themes of Big Data Analytics and Human-Centered Computing. The backbone of Tier 1 is the Department s six expertise areas, namely Big Data Analytics and Information Retrieval, Graphics, Visualization and Multimedia, Human-Centered Computing, Networking and Mobile Computing, Pattern Recognition and Machine Intelligence, and Systems and Software Engineering. Under Tier 2 is Tier 3, which consists of crossarea research centers and laboratories. Jiannong Cao Chair Professor and Head Department of Computing In celebrating the Department s 40 th anniversary and commemorating this important milestone, the Department would like to grasp this golden opportunity to showcase our innovative projects at the Research Open Day to be held on 5 July This brochure in your hands presents these projects which demonstrate the research accomplishments that we have achieved in recent years. We are looking forward to sharing and exploring opportunities of collaboration with you. We recognize the necessity of establishing strong collaboration and partnership with local and overseas institutions and industry partners. We firmly believe that such collaboration and partnership will definitely bring opportunities for generating new ideas and transferring our innovations to the industrial sector, benefiting Hong Kong, the nation and the world. Taking pride in our research undertakings and endeavours over the past four decades, we are looking forward to advancing our success on various fronts in the years ahead through close collaboration and partnership with the academia and industry. Please join me in this exciting venture to continuously strive for excellence in education, scientific discovery and knowledge transfer. 03

5 Research Strengths The Department of Computing is devoted to spearheading top-quality and inter-disciplinary research, keeping up with the knowledge advancement and global development in computing and information technology. The Department has newly inaugurated a 3-tier research framework with a common research infrastructure. The first tier is department-driven strategic research with focused areas of applications. The second and third tiers are departmental research expertise areas and centers and laboratories. Main Research Areas The Department has identified six research areas with high application potentials to drive innovation through smart computing for a better world. (1) Big Data Analytics and Information Retrieval* Big data analytics imposes challenges on many aspects ranging from data acquisition, analysis and visualization to information retrieval and knowledge discovery. The Department tackles both theoretical and application issues in this field, building a big data framework. It applies distributed computing technology to tackle the sheer volume of data and information with consideration of efficiency. It also investigates efficient retrieval of information with respect to semantics and languagespecific factors. (2) Graphics, Visualization and Multimedia With the rapid growth in the size and diversity of multimedia information, the Department researchers focus on some key issues with excellent application potentials. In 3D computer graphics modelling and rendering, content-based image retrieval, and multi-sensor data fusion and motion analysis, the Department applies its output to improve the performance of various applications serving education, entertainment, telemedicine and social networking. (3) Human-Centered Computing Human-centered computing is an emerging interdisciplinary field concerned with computing and computational artifacts as they relate to human behaviour. The Department utilizes computational technologies to analyze humans actions and drive performance enhancement by conducting research in brain informatics, ubiquitous computing, human-computer interface and social computing. 04

6 3-tier Research Framework (4) Networking and Mobile Computing Mobile computing advances with increasing flexibility and freedom as supported by new technologies of wireless networks and smarter mobile devices. The Department works on related areas and extends to mobile cloud computing, mobile applications security, and network performance monitoring and measurement. Such efforts aggregate to bring about enhancements in different application areas including intelligent transport system, structural health care and security. (5) Pattern Recognition and Machine Intelligence In order to align with the challenging trends in the development of pattern recognition and machine intelligence, the Department focuses on biometrics security, medical imaging, video and image processing, and data mining and offers solutions to problems related to bioinformatics, computational biology, web intelligence and social media mining. (6) Systems and Software Engineering The Department works on key challenges in the areas of cyber-physical systems, real-time embedded systems, contextaware middleware, software process and quality, agile development and risk management and it tactfully handles the problems arising from the operation of business critical applications and robust computer-based systems. * Currently departmental strategically focused areas 05

7 Centers and Laboratories The Department currently runs the following state-of-the-art research centers and laboratories to drive and facilitate interdisciplinary and cross-area research of strategic applications. It has also established strong partnership with local and overseas institutions and dominant industry partners with joint laboratories successfully set up and projects of high research and development values undertaken. The Department has established the following centers and laboratories in the past decade: Advanced Enterprise Infrastructure Laboratory / PolyU- SYSU Networked Lab on Creative Biometrics Research Centre Game Laboratory Human Language Technology Laboratory Intelligent Home Internet and Mobile Computing Laboratory Laboratory for Software Development and Management PolyU-IBM Enterprise Social Computing Lab More recently the Department has established the following new cross-area laboratories promoting inter-disciplinary research and applications: PolyU-Yonyou Joint Laboratory on Smart Cloud Computing Advanced Network Monitoring and Measurement Lab Big Data Analytics Lab CHILab Cognitive Computing Lab Embedded Systems and CPS Lab Mobile Cloud Computing Lab Pervasive Visual Computing Lab PolyU Branch of Ministry of Education Key Laboratory of Machine Intelligence and Advanced Computing Smart Healthcare Management Lab Smart Sensing Lab Visual Analytics Laboratory 06

8 Lists of Projects Principal Investigators (in alphabetical order) Professor BACIU George 白求智 BMath; MASc; PhD(Waterloo); MACM; MIEEE; HKDEA 16 VALID 基於 Web 的大規模流數據可視化框架 Research Interest: Interactive Computer Graphics, Virtual Reality, Motion Synthesis, Image and Video Analysis Professor CAO Jiannong 曹建農 BSc(Nanjing); MSc; PhD(Washington State); MACM; SMIEEE; SMCCF 17 A Practical Wireless Structural Health Monitoring System Using Intelligent Sensor Network 基於智能無線傳感器網絡的建築物結構健康監測系統 18 Accelerating Mobile Interactive with a Cloud Computing Platform 基於雲計算的移動應用程式加速服務 19 Cushionware: A Practical Sitting Posture Recognition System Cushionware: 一個實用的坐姿識別系統 20 HAWK: Heterogeneous Advanced Wireless Networks HAWK: 支持安全無縫互聯網接入的先進異構網絡 21 S-Helmet: A Proactive Localization System for Construction Site Management 智能安全帽 : 基於無線實時定位的建築工地安全管理系統 22 Smartphones Safety and Security Services 智能手機安全與保安服務 Research Interest: Parallel and Distributed Computing, Computer Networking, Mobile and Pervasive Computing, Fault Tolerance Middleware Dr. CHAN Chi Fai Stephen 陳志輝 BSc; MSc(Rochester); MSc(Wisconsin-Madison); PhD(Rochester); MIEEE; MACM 23 A Mobile Computing Center: A Lab in a Suitcase on a Tuktuk 知識無界限 : Tuktuk 上的流動學習手提箱 24 i*catch Wearable Computing for Novice and Children 可穿戴智能衣簡易研發系統 25 MelodicBrush 音樂毛筆 : 書法轉化音樂 (Co-Investigator) Research Interest: Data Mining, Recommender System, Human Computer Interaction, Service Learning 07

9 Professor CHAN Chun Chung Keith 陳振冲 BMath; MASc; PhD(Waterloo) 26 An elearning Application using Automated Ontology Management 基於語義搜索引擎的網上學習 27 Big Data Analytics for Drug Design and Discovery 面向新藥研發的生物大數據分析 28 Brain Signal (EEG) Analysis for Real 腦電波分析及其應用技術 29 Community Detection in Social Networks 社交網絡中的社區探索 30 Mining Social Media for Sentiment Analysis 用於用戶情感分析的社交媒體大數據挖掘技術 Research Interest: Software Engineering, Data Mining, Computational Intelligence Dr. CHAN Henry 陳峻斌 BA; MA(Cambridge); PhD(British Columbia); MIEEE 31 RFID-assisted Learning and Computer Interface System for Disabled People 易學習 無障礙電腦學習系統 32 RFID-based Location Tracking System Using a Peer-to-Peer Network Architecture 基於 RFID 點對點網絡架構的定位系統 33 RFIShoe RFID and Bluetooth-based Intelligent Shoe 無線射頻智能鞋 : 腳鼠 操控電腦系統 Research Interest: Networking and Communications, Internet Technologies, Electronic Commerce Dr. CHAN Toong Shoon Alvin 曾東雄 Dip(ECE)(Singapore Poly); BEng(Leeds); PhD(New South Wales); MIEEE; MACM 34 A High-Performance Distributed Publish-Subscribe Architecture 高性能分佈式 發佈 訂閱 架構 Research Interest: Internet and Mobile Computing, Context-aware Computing, Smart Card 08

10 Dr. CHANG Kow Chuen Rocky 張蛟川 BSc(Virginia Poly); MSc; MEng; PhD(Rensselaer); MIEEE; MACM 35 Measuring the Quality of Experience of Residential Broadband Services 家用寬頻服務用戶體驗質量的測量技術 Research Interest: Network Measurement Systems, Internet Infrastructure Security and Privacy; QoE Evaluation of Network Services Dr. CHUNG Fu Lai Korris 鍾富禮 BSc(Manitoba); MPhil; PhD(CUHK); MIEEE 36 Knowledge Transfer for Spectral Clustering 知識轉移的分類技術 37 Large Scale Synchronization-Inspired Partitioning Clustering 基於大數據的分類法 38 Learning Topic Hierarchies for Short Text 更有效的短文文本挖掘 39 Persuasion Driven Influence Propagation in Social Networks 在社交網絡中以勸說驅動傳播的影響 Research Interest: Data Mining, Multimedia, Computational Intelligence, Pattern Recognition Dr. FUNG Sui Leung Walter 馮瑞良 BSc(Iowa State); MSc[PolyU(H.K.)]; MSc(Warwick); EngD(CityU); MIEEE; FBCS; CISA 40 Mobile App for Information Security Awareness Training by 3D Animation 用於改善企業資訊安全的三維動畫智能手機程式 41 Museum of Tea Ware Mobile App with QR Code Identification 茶具文物館智能手機程式 42 Simulation of an Accident & Emergency (A&E) Section of a Hospital Using Flexsim 醫療診所運作的三維模擬 Dr. KUMAR Ajay BEng(Hons); MEng; PhD(HKU); SMIEEE 43 Automated Watchlist Identification System for Enhanced Security at Border Crossings 邊境口岸檢查自動化個人身份識別系統 44 Low-Cost, Contactless and Accurate 3D Fingerprint Identification System 低成本非接觸式精確 3D 指紋識別系統 Research Interest: Biometrics, Computer Vision-Based Industrial Inspection 09

11 Dr. LEONG Hong-va 梁匡華 BSc; MPhil(CUHK); PhD(California); MACM; MIEEE 45 A Physiological Mouse Towards an Emotion-aware Mouse 生理感應滑鼠 Research Interest: Parallel and Distributed Computing, Distributed Databases, Mobile Computing, Internet Computing Dr. LEUNG Kam Nang Hareton 梁金能 BSc(British Columbia); MSc(Simon Fraser); PhD(Alberta); MIEEE; MHKCS 46 idance Human Motion Rhythmic Dance Detection 愛 跳舞 隨時隨地, 心動舞動 47 Smart Health Buddy Instant Image Calorie Counter 健康 友 你 即 照 即知, 熱量無處遁形 Research Interest: Software Testing, Process Engineering and Quality Improvement, ISO 9000, CMM/CMMI, Software Metrics, Software Maintenance, IT Outsourcing Dr. LI Shuai 李帥 BEng(HFUT); MEng(USTC); PhD(Stevens) 48 High Accuracy Control of Multiple Robots for Pollution Localization and Monitoring 多機器人協作式高精度污染源定位與監控 (a participant of the project) Research Interest: Robotics, Dynamic Systems and Control, Recurrent Neural Networks, Distributed Control and Optimization Dr. LI Wenjie Maggie 李文捷 BSc; MSc(Tianjin); PhD(CUHK); MACM; MIEEE 49 Content-Oriented Microblog Mining Based on Speech Act Recognition 基於用戶言語行為的微博內容挖掘研究 50 Modeling Opinion Influence in Social Media Networks for Business Intelligence 社交媒體上的用戶影響力和影響角色分析及其在商業智能的應用 Research Interest: Information Retrieval and Extraction, Natural Language Processing, Temporal Information Processing and Reasoning 10

12 Dr. LIU Yan 劉焱 BEng(Southeast); MSc(Nanjing); PhD(Columbia) 51 What Strikes the String of Your Hearts? 是誰撥動你心弦? 音樂康復系統 Research Interest: Machine Learning, Multimedia Understanding Dr. LO Chi Lik Eric 盧至力 BSc[PolyU(H.K.)]; MSc(HKU); PhD(ETH Zurich) 52 Big Data Analytics as a Service 大數據分析雲端服務 Research Interest: Data Engineering, Software Engineering, Distributed and Parallel Systems Dr. LOU Wei 樓煒 BEng(Tsinghua); MEng(BUPT); PhD((Florida Atlantic); MIEEE 53 Explore a Novel Cross-layer Interference Resistant Multiple Access Approach for WiFi Networks 無線網絡中抗干擾多址接入跨層技術 Research Interest: Mobile Ad Hoc and Sensor Networks, Computer Networks, Mobile Computing, Multimedia Systems Professor LU Qin, MH 陸勤 BSc(Beijing Normal); MSc; PhD(UIUC) 54 Building a Diachronic Chinese Knowledge-Base 歷代漢語知識庫 55 Web Person Name Disambiguation 網路人名歧義消除技術 Research Interest: Natural Language Process and Computational Linguistics, Lexical Semantics, Information Extraction and Knowledge Discovery, Open Systems and Standardization 11

13 Dr. LUI Wing Cheung Richard 雷永祥 BSc(CS); MPhil; PhD (HKU) 56 A Simulation Game for Interactive Learning of Software Project Management 軟件項目管理互動仿真遊戲 Dr. LUK Wing Pong Robert 陸永邦 BSc; Dip Eng(Southampton); MSc(Warwick); PhD(Southampton); CEng.; CITP; SMACM; FBCS; SMIEEE 57 Fast Retrieval with Blind/Relevance Feedback 快速回饋搜查 Research Interest: Information Retrieval, Algorithmics, Pattern Recognition, Speech Processing, Natural Language Processing, Chinese Computing, Computational Linguistics Dr. LUO Xiapu Daniel 羅夏朴 BSc; Msc(Wuhan); PhD[PolyU(H.K.)]; MIEEE 58 NDroid: A New Dynamic Information Flow Tracking Engine for Android NDroid: 針對 Android 應用的新型動態信息流追踪引擎 59 RootGuard RootGuard: Android 手機越獄權限保護系統 Research Interest: Network and System Security, Information Privacy, Internet Measurement, Cloud Computing Dr. NG Hiu Fung Peter 吳曉峰 BSc; MSc; PhD[PolyU(H.K.)] 60 Holographic Display and Interactive Game Development 基於全息投影的互動遊戲開發 12

14 Dr. NG To Yee Vincent 吳道義 BSc(Simon Fraser); MMath(Waterloo); PhD(Simon Fraser) 61 Gen-YOU Net Gen-Y 無毒城 62 spadap-smart Parents Against Drug Abuse Platform 家家齊心抗毒新網絡 Research Interest: Databases, Data Mining and XML, Medical Informatics Dr. NGAI Grace 倪恩恩 ScB(Brown); MSE; PhD(Johns Hopkins) 23 A Mobile Computing Center: A Lab in a Suitcase on a Tuktuk 知識無界限 : Tuktuk 上的流動學習手提箱 24 i*catch Wearable Computing for Novice and Children 可穿戴智能衣簡易研發系統 Research Interest: Human Computer Interaction, Human Centered Computing, Pervasive Computing, Natural Language Processing Dr. SHAO Zili 邵子立 BEng; MEng(Electronic Science & Technology of China); MSc; PhD(Texas at Dallas) 63 Research and Development of a Flash Translation Layer with Demand-based Block-level Address Mapping for Large-Scale NAND Flash Memory Storage Systems 大容量 NAND 閃存的優化技術 Research Interest: Embedded Systems and Real-time Systems Dr. SHIU Chi Keung Simon 邵志強 MSc(City U.K.); MSc(Newcastle U.K.); PhD[PolyU(H.K.)]; MBCS; MACM; MHKCS; MIEEE 64 Augmented Reality Games for Android and ios Platforms 安卓及蘋果平台之增強現實遊戲開發 Research Interest: Soft Computing, Case-based Reasoning, Web 3D Systems and 13

15 Dr. WANG Dan 王丹 BSc(Peking); MSc(Case Western Reserve); PhD(Simon Fraser) 65 ICT Technology for Smart and Green Buildings 綠色智能樓宇中的信息技術 Research Interest: Wireless Sensor Networks, Peer-to-Peer Networks, and QoS Routing Dr. WANG Qixin 王启新 BE; ME(Tsinghua); PhD(UIUC) 66 An Online Risk Confinement, Assessment, and Mitigation Framework for Wireless-Enabled Medical Systems 無線醫療系統的在線風險限制 評估與消減框架研究 67 Google-Glass-Like Helmet for Real-Time Augmented Reality 一種類谷歌眼鏡的即時增強現實頭盔 68 Profiling Faults in Cyber-Physical Systems for Dependability Assurance 面向可靠性保障的信息 物理融合系統故障分析 69 Real-Time Multicast Support in Multi-Hop Switched Industrial Fieldbuses 工業現場總線實時多播機制 Research Interest: Real-time/Embedded Systems, Cyber-physical Systems Dr. XIAO Bin 肖斌 BSc; MSc(Fudan); PhD(UT Dallas) 70 New Approaches to Event Data Collection and Processing by Crowdsourcing to Smartphones 通過眾包到智慧手機的事件與數據收集和處理系統 71 New Technologies for Dynamic RFID Systems: Problems,, and Solutions 動態 RFID 對物體信息的高效收集和處理 72 Wireless Indoor Localization System Design 智能手機用戶的無線室內定位系統 Research Interest: Mobile Cloud Computing, Data Management, Network Security, Wireless Sensor Networks and RFID Systems 14

16 Dr. YIU Man Lung 姚文龍 BEng; PhD(HKU) 73 Automatic Performance Tuning for Spatial Databases 空間數據庫性能調整自動化技術 74 Enhancing Location-Based Services by Using Online Route Services 利用路線信息提升地理位置服務效能 Research Interest: Data Engineering, Query Processing, Spatial Database Systems Professor You Jia Jane 尤佳 BEng(Xi'an Jiaotong); PhD(La Trobe); Dip(Beijing Foreign Language); MIEEE 75 A New Secured Retinal Imaging System for Computer Aided Non-intrusive Diabetic Care 新一代安全可靠非侵入式糖尿病監護的視網膜圖像系統 76 Automated Truck Loading Monitoring for Smart Logistic Management and Road Safety Control 用於智能物流管理與道路安全的車載監控系統 Research Interest: Image Processing, Pattern Recognition, Computer-aided Diagnosis and Monitoring Professor ZHANG Dapeng David 張大鵬 BSc(Peking); MSc; PhD(Harbin IT); PhD(Waterloo); FIEEE; FIAPR 77 Facial Beauty Analysis 人臉美學分析 78 Palmprint Recognition for Security 掌紋識別系統 79 Tongue/Pulse/Odor Detection for Healthcare 健康診斷 : 舌象 / 脈象 / 氣味檢測 Research Interest: Biometrics, Pattern Recognition & Image Processing, Information Security, Neural Networks Dr. ZHANG Lei 張磊 MSc; PhD(Northwestern Polytechnical) 80 Fast Compressive Tracking 基於視頻的高效目標跟踪技術 81 Image and Video Resolution and Quality Enhancement 圖像與視頻分辨率和品質增強系統 82 Low Cost and High Performance 3D Printing System 低成本高性能 3D 打印系统 83 Personal Authentication Using Finger-Knuckle-Print 基於指節紋的身份認證系統 Research Interest: Image and Video Processing, Biometrics, Multisensor Data Fusion, Pattern Recognition 15

17 Professor BACIU George VALID 基於 Web 的大規模流數據可視化框架 Workflow of VALID Client architecture A web framework for visual analytics of big streaming data. VALID is a new web-based framework for the dynamic visualization of big data. The framework is based on the idea that no physical device can ever catch up to the analytical demand and the physical requirements of large data. The platform is based on the traditional server-client model. However, the platform is enhanced by effective analytical methods that operate on data streams, such as binned points and bundling lines that make sense of data reduction and real-time dynamic visualization. Different types of large datasets were used to demonstrate the effectiveness of this framework. High optimized stream generator can support a large number of clients simultaneously for big data processing Stream generator can handle the coming data Dynamic visualization module makes scene of perceptual and animated visualization VALID can be used for visual analytics of large data. For example: Queries critical events from big social network data Observes the airline situation in real-time Visualizes large graph in multi-screen 16 VALID visualizes an airline graph in the United States

18 A Practical Wireless Structural Health Monitoring System Using Intelligent Sensor Network 基於智能無線傳感器網絡的建築物結構健康監測系統 Professor CAO Jiannong Test in Wan Chai Test on Hedong Bridge, Guangzhou Test on CCTV Tower in Beijing A wireless sensor network based system that monitors health conditions of a structure in real-time. The conditions of skyscrapers and long bridges are critical for public safety. Existing systems for monitoring the health condition of civil structures use wired architecture and centralized control, incurring high cost and long deployment time. We have designed a practical structural health monitoring system using intelligent wireless sensor networks. This innovative system can be easily and flexibly deployed on various types of structures, and wireless sensor nodes can automatically organize a network to fulfill the required functions for structural health monitoring. We have tested many real infrastructures, including the CCTV tower in Beijing and Hedong Bridge in Guangzhou and demonstrated that the system can meet the practical requirements of civil engineering. Our technology can also be used to monitor the safety condition of old tenement buildings and structures that might be damaged in a disaster. A practical wireless sensor network based structural health monitoring system that meets the practical requirements of civil engineering Low cost and easy deployment Energy saving by putting sensor nodes to sleep and automatically waking them up when critical event occurs High accuracy of wireless synchronization On-board storage User-friendliness with well-designed GUI showing the health condition of the structure under monitoring, as well as the status of all wireless sensor nodes and allowing the configuration of the parameters of the sensor nodes such as the sampling frequency and duration, etc. Monitors high buildings, long bridges and old buildings and reports the condition when danger occurs Short-term data collection for evaluation purposes Award(s) Special Mention Award and the Certificate of Merit of Hong Kong ICT Awards

19 Accelerating Mobile Interactive with a Cloud Computing Platform 基於雲計算的移動應用程式加速服務 Professor CAO Jiannong csjcao@comp.polyu.edu.hk Touch-free Human Computer Interaction: hand gesture and eyeball movements A cloud-based platform that accelerates mobile interactive applications and saves battery of mobile devices. Mobile augmented reality Modern mobile devices equipped with a multitude of sensors and multimedia capabilities provide opportunities to create a wide range of new mobile applications. However, these applications usually suffer from subpar performance and poor user experiences due to their inherent resource constraints. We have developed a cloud-based platform to accelerate the execution of mobile interactive applications, meanwhile to save the energy of mobile devices. Compared with traditional tools and techniques, we aim to provide Acceleration as a Service (AaaS) with a Cloud Computing platform for the developers. Performance improvement of mobile interactive applications by adaptive partitioning of computation tasks Resources consumption saving at both mobile and cloud sides Easy development and deployment through providing Acceleration as a Service (AaaS) for developers Touch-free Human Computer Interaction on mobile devices: to enable the control of mobile applications by contactless interactive ways, e.g. hand gestures, eyeballs movements, etc. Mobile augmented reality: to recognize the physical environment surrounding the mobile users, and rendering related useful information to users Funding Parties IBM Shared University Research Program RGC General Research Fund 18

20 Cushionware: A Practical Sitting Posture Recognition System Cushionware: 一個實用的坐姿識別系統 Professor CAO Jiannong csjcao@comp.polyu.edu.hk Cushionware Pressure sensors Application: changing sitting posture for playing a game. A system that provides real-time sitting posture recognition and enables posture-based interaction. We have developed a sitting posture recognition system using low cost embedded pressure sensors. The system, called Cushionware, is accurate, non-intrusive and inexpensive. In particular, the system is able to accurately infer a set of common sitting postures based on a number of pressure sensors deployed on the seat cushion. Cushionware can effectively enable users to interact with systems through changing sitting postures. In addition, it is able to provide feedback to users when poor sitting postures such as slouching is detected. High accuracy: currently the average accuracy of sitting posture recognition is 84.7% Non-intrusiveness: the pressure sensors deployed within the cushion are mm thin Low cost: the system components only cost less than HKD1,500 Human Computer Interaction: to enable sitting posture as a novel interaction interface Healthcare: to help promote and improve the healthy sitting behavior 19

21 HAWK: Heterogeneous Advanced Wireless Networks HAWK: 支持安全無縫互聯網接入的先進異構網絡 Professor CAO Jiannong A heterogeneous wireless networking technology for seamless communication and mobility. We have developed a platform of heterogeneous advanced wireless networks known as HAWK, which delivers ubiquitous mobility and coverage for Internet access over heterogeneous wireless networks including 3G, WiFi and mesh networks. HAWK supports fast and seamless handoff across different networks and provides context-aware multimedia service by adapting quality to network conditions. With HAWK, mobile users can enjoy their online games and live broadcast (e.g World Cup) seamlessly and continuously. HAWK also includes custom-made wireless mesh routers that adopt multi-radio multi-channel techniques and optimized routing protocols to achieve ad hoc wireless networking with applications like underground mines and wireless video wireless video surveillance. Seamless roaming among heterogeneous networks Context-aware and adaptive middleware: HAWK is able to adapt to both the change of network and the change of mobile device to provide high quality services for users High system throughput: HAWK adopts multiple-radio-multiple-channel techniques to achieve high system throughput Heterogeneous wireless access networks for ubiquitous Internet coverage Develops the underground wireless mesh networks to support audio and video communications (in collaboration with Landong Laser Technology Ltd. in Shandong China) Award Bronze Medal, Brussels Innova 2009 Wireless mesh router with camera developed by research group System architecture of HAWK 20 Patent No. & Country A Fingerprint Method for Intelligent Handoff Based on Fuzzy-logic. (Application Pending, No , China) A Handoff Mechanism for Wireless Mesh Networks (Patent No , China) A Scheduling Method of Dual-radio Fast Handoff in Wireless Networks (Patent No , China) 一種分組調度方法及系統 ( 中國申請號 : ) 一種預認證和預配置方法及其係統 ( 中國申請號 : ) 感知網絡計費方法 系統及服務器器 ( 中國申請號 : ) Funding Parties Huawei Technologies, Co. Ltd. Innovation and Technology Fund Tsinghua Tong Fang Company Limited

22 S-Helmet: A Proactive Localization System for Construction Site Management 智能安全帽 : 基於無線實時定位的建築工地安全管理系統 Professor CAO Jiannong csjcao@comp.polyu.edu.hk A wireless localization system that tracks locations of construction workers and issues alarms when necessary. S-Helmet We have developed a system for construction safety, called S-Helmet, based on our wireless real-time localization technique. S-Helmet uses wireless tags attached to the safety helmets of workers and moving objects to track the workers locations with high accuracy and timing requirements. When a worker is moving near dangerous zones (e.g. edges of the working platforms, close to moving objects, electric discharge, etc.), S-Helmet will automatically issue an alarm to the worker as well as the construction site coordinators, effectively preventing accidents from happening. Easy deployment: S-Helmet can be easily deployed in construction sites High accuracy: the localization error of S-Helmet is about 1 meter Real-time localization information: location update every second Safety alarm: issues safety alarm in a real-time manner when necessary Node Hardware Construction site management: to improve the safety of construction workers in construction sites Coal-mine applications: to provide the real-time localization of coal-mine workers Funding Party Construction Industry Council Research Fund Experiment 21

23 Smartphones Safety and Security Services 智能手機安全與保安服務 Professor CAO Jiannong Mobile application and software that provide safe unlock and prevent accidents for smartphone users. We have developed augmented smartphones safety and security services with our novel techniques. We proposed a smartphone application, called WalkSafe, to help prevent smartphone users from falling down when they are approaching walls, staircases, uneven grounds, stumbling blocks, etc. We also proposed a method using eye blinking to unlock the touchscreen of smartphones in a robust way that can avoid attacks arising from surfing and smudge. Detects the abrupt change of distance ahead and prevent smartphone users from accidents in an energy-efficient way Touch-free unlocks of smartphones using eye blinking, which is robust to shoulder surfing and smudge attacks and is touch-free The user-interface for touchscreen unlocking using eye blinking Unlocks of mobile phones Pedestrian safety for smartphone users System architecture for Walk-Safe Using ultrasonic to detect unsafe situation ahead 22

24 A Mobile Computing Center: A Lab in a Suitcase on a Tuktuk 知識無界限 : Tuktuk 上的流動學習手提箱 Dr. NGAI Grace csgngai@comp.polyu.edu.hk (Principal Investigator) Dr. CHAN Chi Fai Stephen csschan@comp.polyu.edu.hk (Co-Investigator) Children using the mobile devices to learn The lab in a suitcase. It consists of a server and router and mobile devices. The server hosts e-resources such as children s Wikipedia, an offline version of Khan Academy, and e-books. The fully outfitted tuktuk ready for action with our lab in a suitcase and a stash of donated library books. A mobile computing center for developing regions. The goal of our project is to provide a viable, sustainable digital inclusion solution to children and communities that are off the power grid, but at the same time, close enough to urbanized infrastructure. There are many such communities around the world that are literally a last mile away from digital inclusion. Our solution blends portability and mobility. We have designed a mobile computer lab, which we call the Lab in a Suitcase. It contains a server with elearning resources, a router, batteries and mobile devices. After the lab has been charged up, it can be brought out to the site, and the batteries will provide 5-6 hours of power to the server and the wireless router, which is enough for one intensive teaching session. A donation of a tuktuk, or a motorized trishaw, adds mobility to the portability, in effect creating a mobile computer lab on wheels. In June 2013, a team from The Hong Kong Polytechnic University deployed the Lab in a Suitcase, taught 150 primary school students digital storytelling and blogging using the new equipment, and trained the staff and volunteers of our collaborating NGO to maintain and use the mobile equipment. Developed by The Hong Kong Polytechnic University students as part of a service learning project Brings e-learning and mobile learning to developing areas Digital inclusion for developing countries Service learning for The Hong Kong Polytechnic University students Award Silver Award at The Hong Kong ICT Awards (HKICTA) 2014: Best Digital Inclusion (Service Stream) Funding Parties Donations by CSL and other generous donors The Hong Kong Polytechnic University Community Service Fund 23

25 i*catch - Wearable Computing for Novice and Children 可穿戴智能衣簡易研發系統 Dr. NGAI Grace csgngai@comp.polyu.edu.hk (Principal Investigator) Dr. CHAN Chi Fai Stephen csschan@comp.polyu.edu.hk (Co-Investigator) The i*catch main processor Three girls demonstrating their wearable jacket i*catch in use for storytelling Wearable computing for novice and children. i*catch is a construction kit for physical and wearable computing that was designed to be scalable, plug-and-play, and to provide support for iterative and exploratory learning. It consists of a standardized construction interface that can be adapted for a wide range of soft textiles or electronic boards, a set of functional components, and an easy-to-use hybrid text-graphical integrated development environment. Our objective is to design an easily usable, manufacturable and extensible construction kit that can be used in a wide range of teaching tasks for a wide variety of student demographic profiles. In this project, we present detailed specifications of our construction kit, and explain some of the major design decisions. Experiences in using the kit in multiple teaching environments, ranging from elementary school to postgraduate, demonstrate that the design objectives have been achieved. Pet minder: i*catch with social network Designed and developed with pedagogical concepts in mind Enables plug-and-play for wearable computing and electronic textiles Facilitates learning through exploration and trial-and-error Facilitates creative teaching of computational concepts to children and students Applied to teaching of a variety of topics from computing outreach to a graduate course on human computer interaction 24

26 MelodicBrush 音樂毛筆 : 書法轉化音樂 Dr. CHAN Chi Fai Stephen csschan@comp.polyu.edu.hk Dr. NGAI Grace csgngai@comp.polyu.edu.hk (Co-Investigators) Cross-model digital art creation for calligraphy and music. MelodicBrush is a novel system that connects two ancient art forms: Chinese inkbrush calligraphy and Chinese music. Our system uses vision-based techniques to create a digitized ink-brush calligraphic writing surface with enhanced interaction functionalities. The music generation combines cross-modal strokenote mapping and statistical language modeling techniques into a hybrid model that generates music as a real-time, auditory response and feedback to the user s calligraphic strokes. Our system is in fact a new cross-modal musical system that endows the ancient art of calligraphy writing with a novel auditory representation to provide the users with a natural and novel artistic experience. Experimental evaluations with real users suggest that MelodicBrush is intuitive and realistic, and can also be easily used to exercise creativity and support art generation. MelodicBrush in action: calligraphy is transformed into music notes that are generated in real time. Translates calligraphic strokes into music Links the act of writing with an auditory representation Facilitates a novel artistic experience Increases outreach and awareness of traditional arts Enhances understanding of traditional arts 25

27 An elearning Application Using Automated Ontology Management 基於語義搜索引擎的網上學習 Professor CHAN Chun Chung Keith A universal ontology encompassing the entire knowledge of the world is being built to enable global knowledge sharing and ultimately lead to the transformation into a knowledge society. elearning portal homepage Concept search results search term knowledge Concept map search term knowledge Our project attempts to build an automated universal ontology generator for most knowledge domains in the world. Based on it, a general elearning portal that is powered by this universal ontology generator will be developed. Users can upload thousands of documents to the portal and the system will automatically organize the documents according to its semantic contents. Both keyword-based search and exploratory search will be provided in order to widen the users learning scope and to get them to learn more in-depth. The retrieved documents will be ranked according to closeness of semantic relationship. Document concept discovery algorithm - Discovers concepts in a corpus of documents using advanced data mining techniques Ontology map viewer - Visualizes concepts in the form of ontology map Real-time topic-focused crawler - Extracts contents from Wikipedia in real time Tagging and ranking engine - Tags and ranks documents automatically Knowledge management - Enhances the performance of capturing, discovering, sharing and using organizational knowledge elearning system - Fosters exploratory learning by providing an ontology map with links of related concepts Semantic search - Improves search accuracy by understanding searchers intent and the contextual meaning of terms Funding Party Innovation and Technology Fund 26

28 Big Data Analytics for Drug Design and Discovery 面向新藥研發的生物大數據分析 Professor CHAN Chun Chung Keith A novel approach for Big Data Analytics for applications in drug design and drug side effect predictions. An unprecedented explosion in the amount of data generated at every stage of drug design and discovery has been witnessed in the past few years. The objective of our project is to investigate the possible opportunities associated with the substantial rise in data volume due to the increasing data sources. Being able to cope with the challenges of integrating big data from various sources such as Next-Generation Sequencing, virtual compounds and clinical notes, and of uncovering hidden patterns in such big data can break critical bottlenecks and lead to potential huge impact on humankind. The candidate drugs were assigned to potential side-effects according to drug-protein interactions Availability of data makes computational approaches possible Evolutionary approach to evolve design Cloud computing and big data analytics techniques to evaluate feasibility of design Algorithm substantiates the relationship between drug structure and side effects Potential for new drug development with minimal cost and time Screens large-scale candidate compounds and predicts drug target interactions in the initial phase Predicts drug side-effects and repurposes approved drug based on massive clinical notes in the next phase Implements next generation sequencing for personalized medicine in the third phrase Big Data Analytics to cope with different aspects of drug design and discovery 27

29 Brain Signal (EEG) Analysis for Real 腦電波分析及其應用技術 Professor CHAN Chun Chung Keith Analyzing multiple-channel brain signals (Electroencephalography data) to differentiate a person s different mental states. Our brain signals can be obtained by placing electrodes at various positions of the scalp. Multi-channel Electroencephalography (EEG) signals have much more discrimination information than one single channel can provide. Therefore, we have proposed a novel method to classify brain signal data to help predict a person s mental state. Our technique can be applied to such diverse areas as medicine and human computer interface. For example, we have used it for music therapy to help patients recovering from psychosis. We also plan to apply this technique to smart home control using brain signals. EEG Cap 1 Deals with multi-channel brain signal data Studies the relationship between the change of brain signals and subject s mental state. Applies prediction and classification results to different aspects such as medicine and Human-computer Interaction Music therapy Smart home control Accessible computing Human-computer Interaction Medical diagnosis EEG Cap 2 EEG Analysis Tool 28

30 Community Detection in Social Networks 社交網絡中的社區探索 Professor CHAN Chun Chung Keith Results obtained from three social networks and different communities are with different colors Identifying people with similar needs, interests, habits and characteristics for users to participate in a community or for a company to identify targets for marketing. With the development of an evolutionary algorithm, we have designed an effective technique to help users identify a particular community that they can join to either share what they have in common with the others, or to learn from and know more about the others that are within the same community. Our technique can also be used to identify communities that share common interests, habits, political views, etc. for target marketing. Unlike other techniques for discovering communities in social networks, the evolutionary algorithm that we have developed takes into account not just network topologies, but it also considers the attributes that each member in a social network shares. As a result, users in the same community are not just connected more frequently with each other, but they also share similar characteristics. Discovers more than one community to which each person can belong Discovers community of people that share common attributes The evolutionary algorithm can be implemented on parallel machines to allow community detection of a very large network to be discovered in a very short time Identification of community structure of complex networks Social network analysis Target marketing, advertising and promotion in social networking service sites Automatic recommendation to users to join communities in social networks 29

31 Mining Social Media for Sentiment Analysis 用於用戶情感分析的社交媒體大數據挖掘技術 Professor CHAN Chun Chung Keith Sentiment analysis for UBS hedge fund A sentiment analysis system that can be potentially used for various applications for marketing, political analysis, product development, etc. The pervasive use of social media has generated unprecedented amounts of social data. Social media provides a readily accessible platform for users to share information. Mining social media has its great potential to extract actionable patterns that can be beneficial for business, users and consumers. Social media data are vast, noisy, unstructured and dynamic in nature, giving rise to novel challenges. We have developed a novel sentiment analysis system to discover hidden patterns in social media for product sentiment analysis. Sentiment analysis for the US listed company A library of classification rules - This is not as simple as merely a list of good and bad words. The rules are created using Natural Language Processing techniques and can be used to classify search terms into various sentiment categories A fuzzy logic engine - Extracts hidden association patterns from the ambiguous textual data Marketing - Monitors the reputation of a corporate brand or a product Finance - There are numerous news items, articles, blogs and tweets about each public company. Our system can use these various sources to find relevant articles involving the companies and aggregate the sentiment towards them as a single score, which can then be used by an automated trading system Education - Understands students learning experience such as opinions, feelings and concerns about their learning process Sentiment analysis results 30

32 RFID-assisted Learning and Computer Interface System for Disabled People 易學習 無障礙電腦學習系統 Dr. CHAN Henry Receipt of Gold Award (Most Innovative EPC/RFID Application) at the Hong Kong U-21 RFID Awards 2012 from Miss Janet Wong, Commissioner for Innovation and Technology of the Hong Kong SAR Government Students and Prof. Timothy Tong, President of The Hong Kong Polytechnic University at the PolyU Micro Fund Awards Presentation Ceremony 2012 Enabling disabled students to easily interact with computer systems. System testing Students: LAU Shiu Fung and LO Ka Yee Remark: The aforementioned project information/ description is based on the content of the Hong Kong RFID Awards 2012 Casebook of GS1 Hong Kong: casebooks/rfidawardcasebook_2012.pdf, page 39 People with disabilities in Hong Kong have largely unmet needs and inadequate support when pursuing academic learning. Although many educational institutions provide facilities and services for disabled students, many of these facilities are not completely suitable for people with severe disability. Our ECLearn concept project was designed to address this challenge. It is an RFID-assisted learning and computer interface system for disabled people designed to provide disabled people with upper limb physical handicaps and visual impairments with a barrierfree learning environment. Contains human computer interface devices customized for people with disabilities and support for teachers who are working with disabled students Uses RFID-enabled smart ID cards to identify students with disabilities and store their special requirements Customized computer systems for disabled users Education games for disabled students Award Gold Award (Most Innovative EPC/RFID Application) at the Hong Kong U-21 RFID Awards

33 RFID-based Location Tracking System Using a Peer-to-Peer Network Architecture 基於 RFID 點對點網絡架構的定位系統 Dr. CHAN Henry cshchan@comp.polyu.edu.hk Attaching active RFID tags to objects for location tracking purposes. RFID-assisted location tracking has been put to a variety of commercial applications. One of them is assets management system, where RFID tags are attached to objects to track their movement from place to place and prevent unauthorized action in the process. Our RFID-based location tracking system adopts a peer-to-peer (P2P) network architecture. It allows active RFID readers to communicate and transfer data effectively, hence simplifying the installation effort. Fewer reference tags are needed for calibrating and even so, no complex calibration such as fingerprinting is involved. These tags communicate with RFID readers using a wireless link protocol. The P2P network can be expanded easily as it allows each reader to be placed flexibly on an ad hoc basis. These RFID readers can work collaboratively over the P2P network to estimate an object s location. As this system allows implementation time to be shortened providing a low-cost solution for location tracking accuracy, there is plenty of potential for its application in commercial operations. Bronze Award (Most Innovative EPC/RFID Application) at the Hong Kong U-21 RFID Awards 2013 Use of active RFID Use of a P2P network architecture Asset management Customer relationship management Awards Best Student Paper Award at The 2014 IAENG International Conference on Computer Science Bronze Award (Most Innovative EPC/RFID Application) at the Hong Kong U-21 RFID Awards 2013 Funding Parties Hong Kong RFID Innovation and Technology Fund Experiment by the student Student: HUI Chun Pan Remark: The aforementioned project information/ description is based on the content of the Hong Kong RFID Awards 2013 Casebook of GS1 Hong Kong: casebooks/rfidawardcasebook_2013.pdf, page 39 Basic operation 32

34 RFIShoe - RFID and Bluetooth-based Intelligent Shoes 無線射頻智能鞋 : 腳鼠 操控電腦系統 Dr. CHAN Henry cshchan@comp.polyu.edu.hk Foot mouse functions for people with upper limb disabilities. People with upper limb disabilities cannot use the computer mouse and are therefore deprived of online access. To help them overcome this handicap, we have developed an RFIShoe application to provide foot mouse functions through a pair of special shoes. Gold Award (Most Innovative EPC/RFID Application) at the Hong Kong U-21 RFID Awards 2013 Carrying a RFID tag, a disabled user wearing the customized shoes can set up the mouse and load their personal profiles to the computer automatically. Wherever they go, they can load the same customized setting on different computers. Sensors embedded in the shoes can collect pressure data, through which different foot gestures can be detected so that the computer cursor can be controlled easily. In the default setting, the left shoe is primarily designed for clicking while the right shoe, for moving the mouse in four directions. Hence, without the need for hands, users can move the cursor up just by stepping the top of their right shoe sole. They can also do likewise to move the cursor in other directions. To change their personal settings, they only need to have their preferences added to the RFID tag, including the tasks to be performed by each shoe and the settings of pressure levels. As the shoes draw on RFID technology to complete auto login, their wearers need not input anything by hand. They only need to move closer to a reader for auto login. Furthermore, Bluetooth technology application enables foot pressure data uploaded to a file, where it can be read, analyzed and converted into cursor behaviors. Sets up the mouse and load personal profiles to the computer automatically by carrying a RFID tag Sensors embedded in the shoes can collect pressure data, through which different foot gestures can be detected so that the computer cursor can be controlled easily Students and RFIShoe Students: LIU Xi, SUN Ruo Qing and ZHU Jia Chen Remark/Acknowledgment: 1) The aforementioned project information/ description is based on the content of the Hong Kong RFID Awards 2013 Casebook of GS1 Hong Kong: casebooks/rfidawardcasebook_2013.pdf, page 38 2) The RFIShoe application was developed based on a special shoe system, called Intelligent Footwear System provided by the Institute of Textiles and Clothing of The Hong Kong Polytechnic University. Practical and user-friendly: the RFID-assisted shoes can help eliminate the digital divide for those who have lost the ability for upper limb usage Award Gold Award (Most Innovative EPC/RFID Application) at the Hong Kong U-21 RFID Awards

35 A High-Performance Distributed Publish-Subscribe Architecture 高性能分佈式 發佈 訂閱 架構 Dr. CHAN Toong Shoon Alvin Large-scale dissemination of Extensible Markup Language (XML) structured information based on efficient content-based routing over publish-subscribe middleware architecture. An Extensible Markup Language (XML) based pub-sub system mainly focuses on active, selective and asynchronous dissemination of timely, personalized and dynamic information represented and modeled by XML. Existing approaches in XMLbased pub-sub mainly target at small-to-medium scale and static network, therefore they do not concern about issues such as scalability, robustness and fault-tolerance, leading to great performance penalty when being applied to a wide-area and dynamic network environment. To tackle these issues, we have adopted a series of novel approaches to enable an efficient distributed system design scaling up to operate in a large-scale and dynamic network environment. Efficient event representation and event matching algorithm utilizing bloom filter based matching algorithm A robust, adaptable, yet easy-to-maintain application layer network routing scheme - Self-organizing capability for the overlay infrastructure built on a novel structure-aware key-based routing mechanism - Fault-tolerant architecture that tackles dynamism of the underlying network An extensible multilayer architecture Event-matching framework Push technology Home automation E-commerce E-Health Internet of Things Web-publishing/Social Media Personal Communication/Social Network Funding Party Interdisciplinary/Collaborative Research Grant funded by The Hong Kong Polytechnic University Structure-aware addressing 34 Demonstration of multi-layer event-dissemination

36 Measuring the Quality of Experience of Residential Broadband Services 家用寬頻服務用戶體驗質量的測量技術 Dr. CHANG Kow Chuen Rocky Collects network performance data from home networks to popular websites Participants can view their network performance results in real time A broadband router allowing end users to gauge the quality of experience provided by their residential broadband provide We have developed measurement appliances for residential broadband users to gauge the quality of experience provided by their residential broadband providers. The appliances integrate routing functions with network measurement functions seamlessly. Therefore, no additional appliance is needed for performing the measurement, thus facilitating a wider and faster deployment. The core network measurement technologies are based on the novel methodologies developed from our team s three Innovation and Technology Fund projects. Users can browse the network performance from their home to remote web servers that they usually visit. Apart from browsing their own measurement data, coordinated measurement could also be performed in distributed homes to locate the sources responsible for network performance problems. This information is crucial for resolving disputes between users and their providers, and sometimes between providers as well. In addition, the data collected from the users can be used for an objective comparison of the service quality provided by the four main broadband service providers in Hong Kong. Equipped with novel network measurement tools for accurately measuring end-to-end network performance for web services Provides web-based frontend for users to view their network performance Uses off-the-shelf broadband WiFi routers as measurement nodes For individual users: users can browse the network performance from users home to remote web servers that they usually visit For group users: users can perform coordinated measurement in distributed homes to locate the sources responsible for network performance problems For the public: the data collected from the users can provide an objective comparison of the service quality provided by the four main broadband service providers in Hong Kong 35

37 Knowledge Transfer for Spectral Clustering 知識轉移的分類技術 Dr. CHUNG Fu Lai Korris Incorporating transfer learning ability into spectral clustering. Transferring knowledge from auxiliary datasets has been proved useful in machine learning tasks. However, its adoption in clustering is still scarce. In our project, we consider the problem of learning from auxiliary unlabeled datasets for spectral clustering and propose a new algorithm called Transfer Spectral Clustering (TSC). It involves not only the data manifold information of the clustering task but also the feature manifold information shared between related clustering tasks. Furthermore, it makes use of co-clustering to achieve and control the knowledge transfer among tasks. As demonstrated by the experimental results, TSC can effectively improve the clustering performance by using auxiliary unlabeled data. The first spectral clustering algorithm with transfer learning ability Enhances clustering performance spam filtering News categorization Image clustering Funding Party Interdisciplinary/Collaborative Research Grant funded by The Hong Polytechnic University A conceptual view of transfer spectral clustering 36

38 Large Scale Synchronization- Inspired Partitioning Clustering 基於大數據的分類法 Dr. CHUNG Fu Lai Korris Incorporating transfer-learning ability into spectral clustering. Synchronization-inspired partitioning clustering algorithm has been recently proposed and has been attracting more and more attentions. In order to circumvent the serious difficulty that its existing version can only be effectively carried out on considerably small/medium datasets, we have proposed a novel large-scale synchronization-inspired partitioning clustering (LSSPC) algorithm. LSSPC first condenses a large-scale dataset into its reduced dataset. Then it finishes clustering adaptively on the obtained reduced dataset. Finally, it finishes clustering by further dealing with the remaining objects in the large-scale dataset so that the outliers and singleton clusters can be captured effectively. The effectiveness of the proposed method for large datasets is theoretically analyzed and experimentally verified. Large scale clustering for large datasets Theoretically proved and empirically demonstrated effectiveness in large-scale clustering applications Image segmentation Large-scale clustering applications Funding Party Central Research Grant of The Hong Kong Polytechnic University for RGC General Research Fund Project Rated 3.5 To estimate the queue waiting time by crowdsourcing to smartphones 37

39 Learning Topic Hierarchies for Short Text 更有效的短文文本挖掘 DR. CHUNG Fu Lai Korris The graphical model of ai-hlda Leveraging the hierarchical taxonomy of short text for effective texting mining. One prevalent and distinctive source of data is called short text, e.g. e-shopping reviews, social network status postings and forum posts. As the document-level word co-occurrence possesses the characteristics of being sparse and noisy, single layer representation may fail to grasp the essential semantics and hence generates ambiguity. Hierarchical taxonomy works well in many applications. However, most existing approaches purely leverage one modality of information, e.g. the text itself. Since the short text datasets usually have a richer set of auxiliary information connected through inner taxonomy, a layered mechanism using supportive information modalities will provide better browsing and interpretation of large datasets. A graphic model called ai-hlda is proposed, with flexible usage of the supplemental/auxiliary tagging Encourages clustering and predictive performances in datasets about course selection analysis and analysis were obtained Course selection analysis analysis Social network post analysis Forum post analysis Topical tree inferred from Enron corpus shorter-length subset using the ai-hlda model 38

40 Persuasion Driven Influence Propagation in Social Networks 在社交網絡中以勸說驅動傳播的影響 Dr. CHUNG Fu Lai Korris Studying the diffusion of influence in online social network from a social persuasion perspective. Influence maximization problem, motivated by viral marketing applications, has received extensive attentions in recent years. Its goal is to find a small set of k seed nodes in a social network that maximize the influence spread under certain influence propagation models. In our project, a Persuasiveness Aware Cascade (PAC) model, which considers social persuasion in influence propagation, is proposed. The user-to-user influence probability is estimated by three types of social persuasion, namely, tie strength, peer conformity and authority. Experiments conducted in real-world social networks suggest that our proposed model is more effective in describing real-world influence propagation than the well-studied propagation models for influence maximization. Use of social persuasion in social influence maximization More effective in describing real-world influence propagation Social news influence maximization Movie rating/review influence maximization Viral marketing Funding Party Innovation and Technology Fund Working principle of Persuasiveness Aware Cascade (PAC) models 39

41 Mobile App for Information Security Awareness Training by 3D Animation 用於改善企業資訊安全的三維動畫智能手機程式 Dr. FUNG Sui Leung Walter Mobile app using 3D animation clips to raise staff s information security awareness, such as non-disclosure of passwords, clean desk policy and locking unattended computer screens. Most office users wrongly paste password in open area, inviting hackers Enterprise information security is as strong as the weakest link, normally caused not by machines but human errors. Training to raise employees level of awareness towards information security is therefore urgently needed. However, information security awareness program is both expensive and ineffective. Our mobile application aims to bring 3D animation for information security awareness running on mobile devices such as Android smartphones. The application can provide a novel training program for staff who normally obtain information security resources through reading manuals or attending training workshops. By using this mobile application, the training program can be conducted at flexible place and time. Training records can also be properly maintained for audit and program enhancement purposes. Flexible training Auditable Low-cost Information security training Risk management training All forms of training using animation with mobile devices Award Champion of Information System Audit and Control Association (ISACA) "2013 Mobile App Competition for University & Post Secondary Students" Unlock screen will put company into many kinds of danger, like infection of virus 40 The users will be assessed their understanding after viewing the animation

42 Museum of Tea Ware Mobile App with QR Code Identification 茶具文物館智能手機程式 Dr. FUNG Sui Leung Walter Museums such as the Tea Ware can adopt mobile app to attract visitors Use of QR code can help visitors to identify artifacts faster and easier A mobile app for visitors to use QR code to identify artifacts in the Museum of Tea Ware. A mobile application with QR code identification has been developed to help visitors to effectively identify the artifacts, generating further textural and multimedia information, when they travel to the Museum of Tea Ware. This application is very useful for museums and other organizations to showcase their exhibits and products to a large number of visitors. Other devices like ibeacon Technology can be introduced in the near future. Creates new customer experience Attracts more visitors and increases their satisfaction Automates showcases in museums and exhibitions Museums Events and Exhibitions 41

43 Simulation of an Accident & Emergency (A&E) Section of a Hospital Using Flexsim 醫療診所運作的三維模擬 Dr. FUNG Sui Leung Walter csslfung@comp.polyu.edu.hk Simulating the various scenarios in an A&E section of a hospital when there are many patients queuing for treatments, which can help hospitals for better resource planning. Hong Kong, like most developed cities, has encountered pressing demands for medical services, particularly in the Accident & Emergency (A&E) section during the surge periods. As a result, the hospitals are under tremendous stress. The majority of patients who are critically ill or injured may not be able to receive timely medical treatment. There were cases that patients passed away while queuing for services for a long time in A&E without timely treatment. Consequently, there have been calls for expansions in hospitals to cope with the significant increase in demands for hospital services. However, the complexity of hospital industry makes it difficult to justify expansions for more space, rooms, equipment, nurses and doctors. By using Flexsim, A&E services can be simulated in computers and presented in 3D graphics. This can effectively help the decision makers to plan properly, for example, whether additional resources may enhance service level at acceptable cost, to fulfill their performance pledges in a health care system. Improves customer/patient services Allows better planning for hospitals' expansion Caters for peak usage (e.g. holidays) Provides professional training for staff Application Hospital Accident & Emergency operations A typical Accident & Emergency operation in a clinic 42

44 Automated Watchlist Identification System for Enhanced Security at Border Crossings 邊境口岸檢查自動化個人身份識別系統 Dr. KUMAR Ajay Automated iris segmentation results for the system being developed at Department of Computing of The Hong Kong Polytechnic University Development of automated biometrics system to identify wanted suspects with make-up, masks and disguise at border crossings. Web-based interface for law enforcement officers to modify and add watchlist and inspection System interface (first version) under evaluation for watchlist matching using visual biometrics Currently available methods of watchlist identification at border crossings are either largely manual or operated under highly constrained settings that require a high degree of cooperation from the users. The use of facial masks and makeup, due to religious belief and practices, have posed new challenges to commonly practiced face-recognition-based watchlist identification. Therefore, the screening process at border crossings is often not accurate and considerably timeconsuming, especially when a large number of subjects are to be examined. Our system will have a wide range of applications ranging from border checkpoints to law enforcement and will provide critical early warnings for the governments to protect citizens against crimes, terrorism and other human-based threats. At-a-distance biometrics matching using iris and region around iris Disguise, make-up and spoof detection alerts for immigration officers Advanced watchlist identification in less constrained environment using multispectral imaging Immigration and border crossings Automated surveillance for high security Search for missing children and wandering elderly Funding Party Innovation and Technology Fund Block diagram for the automated watchlist identification system for enhanced security 43

45 Low-Cost, Contactless and Accurate 3D Fingerprint Identification System 低成本非接觸式精確 3D 指紋識別系統 Dr. KUMAR Ajay Reconstructed 3D fingerprint image using single camera Computing relative localization of a 3D minutiae with a reference minutiae (mr) in a given 3D fingerprint template Development of low-cost and touchless 3D fingerprint identification for higher security and better hygiene. Extended minutiae feature information recovered in 3D space for more reliable fingerprint identification. Traditional fingerprint identification by pressing or rolling finger against the hard surface often results in partial or degraded images due to improper finger placement, skin deformation, slippages and smearing. Therefore, touchless 3D finger imaging can provide more accurate personal identification as rich information is available from 3D fingerprint images. We have developed an advanced biometric touchless 3D fingerprint identification system to provide inexpensive, faster, higher security, more hygienic and accurate identification for a wide range of applications in civilian and forensic. Contactless 3D fingerprint recovery and matching using single camera Extended minutiae representation and matching in 3D space Simultaneous 2D and 3D fingerprint matching for higher accuracy and security Lower cost and superior performance than existing state-of-the-art 3D fingerprint systems Higher security for automated immigration checks at border crossings Touchless fingerprint identification for improved hygiene Automatically detection of surgically altered fingerprints Patent No. & Country Patent Application No , United States Funding Party RGC General Research Fund 44

46 A Physiological Mouse - Towards an Emotion-aware Mouse 生理感應滑鼠 Dr. LEONG Hong-va cshleong@comp.polyu.edu.hk A rough model of our physiological mouse User using the mouse Towards ubiquitous emotion awareness. Human-centered computing is rapidly becoming a major research direction in human-computer interaction research. Among the various research issues, we believe that affective computing, or the ability of computers to react according to what a user feels, is very important. In order to recognize the human affect (feeling), one can rely on the analysis of signal inputs captured by a multitude of means. In our project, we propose to make use of human physiological signals as a new form of modality in determining human affects, in a non-intrusive manner. This is achieved via the physiological mouse, as the first step towards affective computing. We have augmented the mouse with a small optical component for capturing user photoplethysmographic (PPG) signal. With the PPG signal, we are able to compute and derive human physiological signals. We have built a prototype of the physiological mouse and measured raw PPG readings. And we have performed experiments to study the accuracy of our approach in determining human physiological signals from the mouse PPG data. We believe that our research will provide a new dimension for multimodal affective computing research. Use of everyday devices to unobtrusively detect physiological signals Detects user stress and tension when using the computer Captures physiological signals for health condition 45

47 idance - Human Motion Rythmic Dance Detection 愛 跳舞 - 隨時隨地, 心動舞動 Dr. LEUNG Kam Nang Hareton hareton.leung@polyu.edu.hk A system that recognizes and matches dance steps with pre-recorded videos. In view of the ubiquitous use of mobile phones, we worked with a local private company to develop an app with both and PC versions that can recognize and match dance steps with pre-recorded videos. When people are dancing according to the videos (with three difficulty levels for selection) shown on the monitor, the system will capture the users movements and compare them with the default ones. The system will provide scores after the dancing concludes. The users can also control the cursor remotely through hand gestures. Intelligent hand gestures recognition Accurate dance steps recognition User Interface iphone portion Controls cursor remotely through hand gestures Gives scores according to users dance steps in comparison with the default ones Funding Party Innovation and Technology Fund PC version of gesture recognition demo 46

48 Smart Health Buddy - Instant Image Calorie Counter 健康 友 你 - 即 照 即知, 熱量無處遁形 Dr. LEUNG Kam Nang Hareton hareton.leung@polyu.edu.hk Protoype of our app A mobile health app promotes healthy living Owing to the rapid increase in Hong Kong s obese population, ubiquitous use of mobile phones, photo-taking of meals before eating, in addition to posting photos to social networks for sharing, we have developed an app, which recognizes Dim Sum, to integrate the above elements to improve the well-being of our target: health conscious people. Through this app, users are able to take pictures of the food they eat, and the app will automatically provide the users with nutritional value and warnings if needed. The users can also share health information on social networks, whereby they are encouraged by their friends to adopt a more healthy diet. Locals love taking photos of their food Food recognition Interaction and reinforcement from peers via social networks Takes photo of food before eating for analysis Image processing technology to simplify data entry Uses social networks to help stay motivated Gives scores to users weight-related behaviours Recommends new friends according to scores Growth survey Photo courtesy: The Chinese University of Hong Kong and Department of Health, HKSAR Sample social network pages of meals 47

49 High Accuracy Control of Multiple Robots for Pollution Localization and Monitoring 多機器人協作式高精度污染源定位與監控 Dr. LI Shuai Close-loop high performance control of multi-robots for cooperative tasks. The snapshots of the robot movements for a typical simulation run with 30 robots for cooperative plume front tracking About 1.3 million metric tons of Crude oil was spoiled in the Gulf of Mexico years ago and resulted in severe impact to the local ecology lasting for several months with a large area of water polluted. This issue appeals novel strategies to effectively predict and monitor the chemical propagation in face of complex flow dynamics and strong environment uncertainties. Targeting at this issue, we have proposed a strategy to use robots to probe the environment agilely, identify the boundary between polluted and un-polluted areas intelligently and track the propagation accurately by leveraging the cooperation of a robotic team. Our project have established a rigorous control strategy integrating the prior knowledge on environmental information with sensing data for optimal detection of the environment proactively and efficiently. Works for dynamic plume where the plume front propagates along time The algorithm still works well even for the case with several leakage sources Two objectives, i.e. formation and plume front tracking, are reached simultaneously The derived analytic solution avoids computationally expensive numerical process for solving the complex plume dynamics Environment surveillance Pollution monitoring and pollution source localization (e.g. active monitoring of air pollutions, gas leakage localization, etc.) * Dr. LI was a participant of this project during his Ph.D. study. Funding Party US National Science Foundation funding (awarded to Professor GUO Yi, Principal Investigator of this project, at Stevens Institute of Technology) Control diagram for the ith robot for multi-robot plume front tracking 48

50 Content-Oriented Microblog Mining Based on Speech Act Recognition 基於用戶言語行為的微博內容挖掘研究 Dr. LI Wenjie Maggie Automatic Twitter topic summarization with speech acts. A snapshot of #sincewebeinghonest Tweets with speech act annotations With the growth of the social media service of Twitter, automatic summarization of Twitter messages is in urgent need for efficient processing of the massive tweeted information. Unlike the traditional summarization, Twitter topic summarization must handle the numerous, short, dissimilar, and noisy nature of tweets. To address this challenge, we propose a novel speech act-guided summarization approach. Speech acts characterize tweeters communicative behaviors and provide an organized view of their messages. The recognized speech acts in tweets are then used to direct the extraction of key words and phrases to fill in templates specially designed for speech acts. A novel approach using speech acts, which captures the common grounds of tweets from a communicative perspective to guide the summarization process An effective abstractive summarization framework fitting the numerous, short, and jumbled nature of tweets Understands user s communicative behaviours with a topic Applies text summarization techniques to solve the problem of information overload caused by the massive and redundant information available online Speech act distributions in the 6 sample Twitter topics Funding Party National Natural Science Foundation of China Types of speech acts 49

51 Modeling Opinion Influence in Social Media Networks for Business Intelligence 社交媒體上的用戶影響力和影響角色分析及其在商業智能的應用 Dr. LI Wenjie Maggie Influence role detection on social media networks. As one of the most popular social media platforms, Twitter has provided people with an effective way to communicate and interact with each other. It is an undisputable fact that people s influence plays an important role in disseminating information over Twitter network. Though a number of research work on finding influential users has been reported, they never really try to distinguish and analyze different types of influence roles, which are of great value for various marketing purposes. Our project has moved a step forward to detect influence roles of Twitter users who are interested in a certain topic. Social Influence The first work targeting the problem of influence role detection A Constrained Multi-view Influence Role Clustering (CMIRC) approach partitioning influential users into five pre-defined categories Creates new business opportunities with a variety of online marketing activities Helps enterprises to locate favorable influential users and manipulate users opinions to attract potential customers and improve sales Business marketing management Social Influence Roles Funding Party RGC General Research Fund Social Influence Marketing 50

52 What Strikes the String of Your Hearts? 是誰撥動你心弦? - 音樂康復系統 Dr. LIU Yan csyliu@comp.polyu.edu.hk An online personal music therapy heals mental illnesses using information technology. The cover of a Chinese music therapy CD Music therapy is the use of music therapy programs to aid patients physiologically, psychologically and emotionally during their treatment of illnesses. Research findings have shown that musical stimuli activate specific pathways in several brain areas associated with emotional behaviors such as the hypothalamus and prefrontal cortex. We have designed an online personal music therapy solution for persons suffering from mental problems and carried out a pilot study with the University Health Service of The Hong Kong Polytechnic University on patients with chronic nonmalignant pain. Seventy five percent of patients stopped using painkillers after listening to a specially designed Chinese Pipa for four weeks. In addition, their levels of anxiety have been substantially reduced. Personalized design for different patients Effective privacy protection Convenient for frequent usage Low-cost Increasing need for modern society and local community The use of music as a healing intervention has been well documented. Some examples are given below: Psychiatry (e.g. Depression, Schizophrenia and Aphasia) Neurology (e.g. Autism and Dementia) Surgery care (e.g. reduces stress, soothes pain and energizes the body) General geriatric care (e.g. Amnesia, Coronary care and child development) Research team attending a seminar on Chinese music therapy 51

53 Big Data Analytics as a Service 大數據分析雲端服務 Dr. LO Chi Lik Eric ericlo@comp.polyu.edu.hk Providing big data analytics as a cloud service in a costeffective manner. Many big data analytics tools are powerful but difficult to deploy and use. In this project, we study the issue of offering those tools as a service. Users can enjoy the service without being bothered by the technical issues such as set-up and machine administration. Our research focuses on how to offer a reliable and lowcost service that makes better use of the limited resource. Reliable Pay-as-you-go Economical Application Big data analytics System architecture 52

54 Explore a Novel Cross-layer Interference Resistant Multiple Access Approach for WiFi Networks 無線網絡中抗干擾多址接入跨層技術 Dr. LOU Wei csweilou@comp.polyu.edu.hk Elimination of exposed terminal problem using signature detection. Our project aims to design a novel cross layer protocol that can combat exposed terminals and exploit the concurrent transmissions in wireless networks, both assisting in achieving higher throughput compared to the standard and other recent protocols. By observing that nodes in current protocols waste transmission opportunities in two different scenarios, we propose a signature detection method in the physical layer to combat control frames collisions, thus exploiting the concurrency at the transmitter side. We have also designed a new net allocation vector update scheme in the medium access control layer to differentiate the interference ranges of different transmission links, exploiting the concurrency of all non-interfering links. Our proposed protocol is based on the standard, therefore it can be implemented into current devices. Experiment and simulation results have demonstrated the feasibility and outperformance of the signature detection method compared to other protocols. Exploits the concurrent transmission opportunities efficiently Can be implemented into current based devices easily Improves the wireless network throughputs significantly Application Wireless local area network to improve the wireless network throughput A typical application scenario exploiting the concurrent transmissions Funding Party RGC General Research Fund New control frames format (the light gray fields are newly added fields) System architecture (the gray blocks are newly added components) 53

55 Building a Diachronic Chinese Knowledge-Base 歷代漢語知識庫 Professor LU Qin, MH csluqin@comp.polyu.edu.hk Multi-level classification of personal names & terms of address Work process Study of titles based on diachronic corpora using Conditional Random Fields (CRFs) model Establishment of an annotated corpus of Ming-Qing and pre-qin classics as well as corpus-based studies on terms of address in classical Chinese. Named entities, such as personal names, place names and terms of address, are crucial elements in the processing of natural languages. Terms of address serve as essential information carriers in any language, but the current research on Chinese information processing has not set up a classification and annotation scheme for them. We have established a scheme of word segmentation and named entity annotation for Chinese classics, focusing on personal names and terms of address. The scheme has been applied to build a diachronic corpus of four influential novels produced in the Ming and Qing dynasties as well as two classics in pre-qin periods through machine learning algorithms. Based on the corpus, terms of address in Chinese classics have been studied from the perspective of natural language processing. A comprehensive scheme for building language knowledge bases in a computer-aided method A diachronic corpus including Chinese classics of different styles Customized specifications for word segmentation and named entity annotation of Chinese classics Studies on named entities, particularly terms of address, the first of its kind, based on the knowledge base Specifications, process, and tools: provides references for the construction of other diachronic knowledge bases Corpus: can be used for research in the fields of information technology, linguistics, literature and history Funding Party Chiang Ching-kuo Foundation for International Scholarly Exchange 54 Classification of named entities

56 Web Person Name Disambiguation 網路人名歧義消除技術 Professor LU Qin, MH Web Person Disambiguation Flowchart Web person disambiguation using machine learning method based on naturally annotated resources. Searching information about persons on the Internet is one of the most common activities for users. However, a name mention can refer to different named entities in real world. To disambiguate named entities over the Internet, we have designed a hierarchical co-reference resolution system using keyword features, which are extracted by the Conditional Random Fields (CRFs) model using Wikipedia as a naturally annotated resource; we then build up personal profiles for each disambiguated individual with two semantic constraints in a graphical model. This system can be easily deployed after a crawler obtains a set of web documents. It can also be extended to resolve the ambiguity of location or organization names. Web Person Disambiguation State-of the art results on WePS data Person name disambiguation in a hierarchical co-reference structure Personal profile building with two semantic constraints in a graphical model Semantic feature extraction using naturally annotated resources Domain-insensitive and practical State-of-the-art performances Improves search engine functions Personalized web search results Resolves author co-reference or citation matching Personal Profiling State-of the art results on CoNLL and Wikipeida data Semantic Feature Extraction (keyword) State-of-the-art results on Wiki-20, Abstract and SemEval datasets 55

57 A Simulation Game for Interactive Learning of Software Project Management 軟件項目管理互動仿真遊戲 Dr. LUI Wing Cheung Richard Project planning: team member selection and activity assignment Main screen during project execution A project management simulation game for supporting the interactive learning of software project management. Project Management Simulation (PMS) is a project simulation game for supporting the active learning of software project management. The player acts as a project manager to plan for a hypothetical software development project scenario to complete the project on time, within budget, with good quality and achieve good stakeholder satisfaction. In the game, the player can analyze the stakeholders expectation, define the project scope, select appropriate team members and assign them to complete the project activities. The player can also motivate and communicate with the project team to improve productivity, communicate with stakeholders, plan for quality control activities, monitor the project progress and perform appropriate project control activities throughout the project. Engages students in active learning through the use of information technology Focuses on project decisions for communication management, stakeholder management and quality management not well addressed by the existing project management simulation games Can be used to support the teaching and learning in IT project management courses Allows students to apply and test the project management concepts learnt in the course and explore good project management practices not covered in lectures. Planning of management activities Funding Party Teaching Development Grant 56

58 Fast Retrieval with Blind/ Relevance Feedback 快速回饋搜查 Dr. LUK Wing Pong Robert Accelerating blind/relevance feedback using a forward index. Blind Feedback or Relevance Feedback is slow for many search applications because its retrieval time is linearly dependent on the number of different words in the query, and effective queries typically have a sizeable number of query terms. As part of a general approach to reduce retrieval times, we have designed specialized forward indices for subsequent fast retrieval. These specialized indices are for storing document vectors. The trade-off between retrieval effectiveness and retrieval time-space efficiency has been investigated using benchmark data (e.g., using Text Retrieval Conference Data) Uses forward indices for fast re-ranking Uses specialized and patented dictionary data structure for fast matching Little loss of retrieval effectiveness improvement by re-ranking instead of reretrieval Accelerates blind feedback Accelerates relevance feedback Improves (web) search engines retrieval effectiveness and efficiency Patent No. & Country Patent No. US B2, United States Patent No. US B2, United States Pseudo Relevance Feedback run times for Text Retrieval Conference Data-6 queries using various forward index methods 57

59 NDroid: A New Dynamic Information Flow Tracking Engine for Android NDroid: 針對 Android 應用的新型動態信息流追踪引擎 Dr. LUO Xiapu Daniel csxluo@comp.polyu.edu.hk A dynamic taint analysis engine for Android applications using native code. NDroid is a new dynamic taint analysis engine for the increasing number of Android applications that use native codes for better performance or security considerations. By attaching taints to selected sources, NDroid can identify the paths where the information will be propagated, even when the information is delivered to and/or fetched from native codes. The first system conducting dynamic taint analysis for Android applications using native code Detects advanced android malware Facilitates the vulnerability discovery for Android applications Android malware analysis Android application vulnerability discovery Funding Parties Innovation and Technology Fund Personal Research Account funded by Department of Computing of The Hong Kong Polytechnic University 58 NDroid's architecture

60 RootGuard RootGuard: Android 手機越獄權限保護系統 Dr. LUO Xiapu Daniel A system for protecting rooted Android smartphones. Rooting Android smartphones can enhance the capability of the device but also introduce new security threats. Existing root management systems can be easily compromised by malicious applications. RootGuard provides fine-grain control on root privileges and can defend against the attacks from malicious applications with root privileges. Provides fine-grain control on root privileges Defend against the attacks from advanced malicious applications Protects rooted Android smartphones Extensible to detect advanced Android malware Funding Parties Innovation and Technology Fund Personal Research Account funded by Department of Computing of The Hong Kong Polytechnic University The RFID system with reader and active tags 59

61 Holographic Display and Interactive Game Development 基於全息投影的互動遊戲開發 Dr. NG Hiu Fung Peter Raising gaming experience above and beyond. We have developed different holographic display prototypes by using Kinect, Leap Motion and PlayStation Vista to explore and test new gaming experience. Holographic display and related technologies push boundaries and expectations for interactive game development. Through integrating with different sensors, they provide a new Human-Computer Interaction experience to captivate the audiences with a real-time illusion and a lasting impression. There are wide applications for holographic display ranging from entertainment, teleconferencing to product showcases in corporate events. Provides a novel gaming experience Raises gaming experience far exceeds the common and mundane expectations Delivers a truly captivating experience at different events Entertainment Teleconferencing Showcases and Events Gifts and interactive toys Holographic display for a handheld device Holographic display with Leap Motion Holographic display showcasing Department of Computing 60

62 Gen-YOU Net Gen-Y 無毒城 Dr. NG To Yee Vincent Our project aims to develop effective digital communication with our young students as well as having new lives (avatars) in an online 3D virtual world for drug prevention. It includes two major components. The first component is a web portal which serves as a common portal for students, teachers, social workers, policeman and NGOs. The second component is an online 3D Social Networking Community (called AD-Island) for students, social workers, mentors and volunteers to interact in a virtual world. The project is targeted at students aged from 10 to 18 and supported by Police Community Relation Office in Tai Po District. An overview of AD-Island Gen-YOU Net 5-zone Island - Red Zone: simulations and games - Green Zone: exhibition areas - White Zone: a Crime Information Centre - Blue Zone: seminar facilities - Yellow Zone: virtual artifacts workplace A 3D virtual world with games and exhibitions containing anti-drug message A learning area for school students to build 3D artifacts and simple programming An online counseling service for the young people in need Funding Party Beat Drug Fund Places in the AD-Island 61

63 spadap-smart Parents Against Drug Abuse Platform 家家齊心抗毒新網絡 Dr. NG To Yee Vincent Location search News center Our district-based project aims at alerting parents and their children of potential drug abusers in their families and teaching youths about the serious consequences of drug abuse. The main activities include developing a mobile community of practice platform for parents to access timely anti-drug information, to have an intelligent alerting tool for detecting potential family drug abusers and to form a mutual support community between parents and other stakeholders such as social workers, anti-drug experts and teachers. In addition, this project enables us to disseminate anti-drug messages to the general public through booklets, videos and mobile platforms. An intelligent alerting tool utilizing artificial intelligence technologies with a mobile front end for assisting parents in possible early detection if there is a drug abuse case in the family Knowledge repository and information help centre for giving advice on how to manage the young family member with potential drug abuse A platform for detecting potential drug abuse cases and a parent community A repository containing online information for parents in handling family drug abuse cases Publishing an electronic booklet including student works to disseminate antidrug information Funding Party Beat Drug Fund 62

64 Research and Development of a Flash Translation Layer with Demandbased Block-level Address Mapping for Large-Scale NAND Flash Memory Storage Systems 大容量 NAND 閃存的優化技術 Dr. SHAO Zili cszlshao@comp.polyu.edu.hk A flash translation layer with demand-based block-level address mapping for large-scale NAND flash memory storage systems. NAND flash memory has been widely adopted in the consumer electronics industry. The increasing capacity of NAND flash memory leads to large RAM (main memory) footprint on address mapping in the Flash Translation Layer (FTL) that is a flash management software module embedded inside a flash memory system. Our project aims to design and implement a novel FTL with demand-based block-level address mapping for large-scale NAND flash. The objective is to reduce RAM footprint without sacrificing too much system response time. Through our novel design, we can considerably reduce the cost of a product with NAND flash memory. Special Features and Advantages Increases the capacity of flash memory dramatically Reduces the RAM footprint in flash memory management Reduces the RAM requirement with very little performance overhead to substantially lower the product cost with flash memory Storage systems Consumer electronics devices Patent No. & Country Effective Caching for Demandbased Flash Translation Layers in Large-Scale Flash Memory Storage Systems (Patent Application No , United States) Funding Party Innovation and Technology Fund Flash Translation Layer (FTL) and Memory Technology Device (MTD) System architecture System architecture 63

65 Augmented Reality Games for Android and ios Platforms 安卓及蘋果平台之增強現實遊戲開發 Dr. SHIU Chi Keung Simon Augmented Reality creating an enhanced experience for users to interact with virtual objects in a familiar environment. Augmented reality games use real surroundings and create game objects within the environment and allow the players to operate and interact. Players feel like within the frame of the camera when pointing at it to where they want the virtual objects to exist and play with. The players can race a car, solve puzzles, place virtual objects, visit a museum and even read a book with magical effects from time to time. Mobile augmented reality for Android and ios platforms is a relatively new technology, and with large potential applications in tourism, education, entertainment and media. By combining virtual data with the physical real world as well as Android and iphone s compass, camera, Global Positioning System system and the like, the resulting user experience will be beyond the usual time and space limitations. A software platform for the best and most creative augmented reality experiences An augmented reality system for mobile phones that matches camera-phone images against virtual objects A game play design method for augmented reality systems Education and training for industrial maintenance service Location-based services with mobile augmented reality systems Wearable human-computer interaction Enhanced Global Positioning System and navigation applications Sightseeing and tourism industries Medical and surgery education and training Funding Party Departmental Earnings funded by Department of Computing of The Hong Kong Polytechnic University 64 An augmented reality game

66 ICT Technology for Smart and Green Buildings 綠色智能樓宇中的信息技術 Dr. WANG Dan A mobile app for fine control of the heating ventilation and air-conditioning system An occupant-participatory approach for thermal comfort enhancement and energy conservation in buildings. Commercial building is one of the major energy consumers worldwide and the heating ventilation and air-conditioning system dominates the total energy consumption. To minimize complaints, the current practice of facility management is to adopt relatively conservative temperatures, leading to massive waste of energy. In this project, we have actively taken thermal comfort into consideration. We propose a participatory approach allowing the occupants to provide feedback regarding their comfort levels. Our novel design reduces intrusiveness of the system by developing a model that can build a profile for each occupant. The decision of set-point temperature is primarily model-driven. We have implemented our design and conducted field experiments at The Hong Kong Polytechnic University and a commercial office in Cyberport. Results showed that our algorithm can successfully maintain high thermal comfort, while reducing 18% of energy consumption. Non-intrusiveness High thermal comfort Energy conservation Application Commercial offices in high-end buildings Experiments in Cyberport Funding Parties Internal Competitive Research Grant funded by RGC Direct Allocation of The Hong Kong Polytechnic University National Natural Science Foundation of China RGC General Research Fund 65

67 An Online Risk Confinement, Assessment, and Mitigation Framework for Wireless- Enabled Medical Systems 無線醫療系統的在線風險限制 評估與消減框架研究 Dr. WANG Qixin A framework of technologies for supporting wireless medical cyber-physical systems dependability. Conceptual framework overview Medical Cyber-Physical Systems (CPS), systems that seamlessly integrate computerized medical devices, medical procedures, healthcare personnel and patients, have received much attention recently. One inevitable trend in medical CPS is the increasing use of wireless-enabled medical devices. However, new challenge may arise from complex communications between system components and environment over wireless channels, as wireless communications tend to be unreliable. To address this new challenge, we propose to build a novel framework to confine, assess and mitigate the risks of wireless failures in medical CPS. Software defined radio Wireless body area networks Medical device plug-and-play Operation room of the future Medical device plug-and-play Wireless medical settings Funding Party RGC Early Career Scheme Experiment 66 System architecture

68 Google-Glass-Like Helmet for Real-Time Augmented Reality 一種類谷歌眼鏡的即時增強現實頭盔 Dr. WANG Qixin A Google-glass-like helmet for real-time augmented reality applicable for life-threatening and critical situations. Augmented reality has broad applications ranging from entertainment, assisted living, to life-threatening and critical scenarios such as surgery and construction site safety. We have designed a Google-glass-like helmet with 3D printing and implemented preliminary real-time scheduling in the Operating System Kernel. The Google-glass-like device designed for critical life situations is different from those for ordinary (e.g. entertainment) applications. Its unique hard real-time support necessitates sophisticated Operating System and distributed systems research and support. For example, in surgery, a Google-glass-like device can clearly display the delicate blood vessels and nerves beneath the skin of the organ that the surgeon is about to dissect, thus preventing possible accidents. In construction sites, a Google-glass-like device can highlight the dangerous objects above the head of a worker, alarming him/her to stay away from dangerous zones. Future research can be directed towards re-designing various augmented reality algorithms that exploits cloud computational power to overcome the resource constraints of the Google-glass-like helmet. The outlook of Google-glass-like helmet Hard real-time augmented reality support Computing workload balance between extremely resource constrained mobile device and cloud 3D printing Real-time augmented reality for surgery safety Real-time augmented reality for construction site safety Funding Party Personal Research Account funded by the Department of Computing of The Hong Kong Polytechnic University The view from Google-glass-like helmet 67

69 Profiling Faults in Cyber-Physical Systems for Dependability Assurance 面向可靠性保障的信息 - 物理融合系統故障分析 Dr. WANG Qixin csqwang@comp.polyu.edu.hk A methodology for profiling the fault propagation across domain boundaries in controlled cyberphysical systems using interdisciplinary knowledge. The integration of computers with the physical world results in Cyber-Physical Systems (CPS), where CPS applications are becoming increasingly critical for life-threatening situations. However, the CPS environment is challenged by the existence of faults/perturbations in this complicated multi-domain physical and cyber world. Although fault propagation in pure cybersystems is a well-studied problem, the challenge of profiling CPS fault propagation lies in how to quantify faults across domain boundaries. Our project focuses on developing a novel framework for addressing the cross-domain issue with a rigorous and analytical perspective and adopting a hybrid modeling, an emerging powerful tool, for CPS research. Inter disciplinary Cross domain boundary Hybrid model checking Cyber-physical systems dependability Cyber-physical systems risk analysis Funding Party RGC Germany/HK Joint Research Scheme A cross-cyber-physical-boundary error propagation profiling framework 68

70 Real-Time Multicast Support in Multi-Hop Switched Industrial Fieldbuses 工業現場總線實時多播機制 Dr. WANG Qixin A real-time switch architecture for supporting multi-hop switched industrial fieldbuses for control applications. Industrial fieldbus based multi-input and multi-output (MIMO) control is a key application in Cyber-Physical Systems research. Due to the rapid expansion of MIMO control systems, fieldbuses have to evolve toward multi-hop switched networks. This implies the traditional shared medium based real-time multicast no longer works. On the other hand, real-time multicast is indispensable for MIMO control. We have noticed that the multi-hop switched fieldbus multicast is different from Internet multicast, due to real-time requirement, co-existing periodicity and synchronicity, unique traffic patterns and confined extent. In this project, we plan to utilize all these features to support real-time multicast in multi-hop switched fieldbuses. Hard real-time support Crossbar Switch Fieldbus Industrial control Avionics Funding Party Newly Recruited Junior Academic Staff funded by The Hong Kong Polytechnic University Object representation and operations 69

71 New Approaches to Event Data Collection and Processing by Crowdsourcing to Smartphones 通過眾包到智慧手機的事件與數據收集和處理系統 Dr. XIAO Bin To monitor the moving speed in an overcrowded footbridge To forecast the crowd density in a canteen during the peak hours To estimate the queue waiting time by crowdsourcing to smartphones New approaches using smartphones for efficient event data collection and processing for smartphone users to timely obtain public event information. Our project investigates a new problem of how to collect and process event data by means of crowdsourcing to event participants smartphones. The event under study is confined to an area, e.g. a canteen, a podium, a seminar room, a footbridge, etc. Free Wi-Fi Access Points (APs) can cover those areas and participants are connected via their smartphones. Solving this problem remains to be technically challenging and none of the previous projects have addressed this problem. Our project goal is to build a crowdsourcing platform, which consists of a cloud server, multiple APs and a large number of smartphones. APs are pre-deployed to provide free wireless services and localization coordinates. Through our platform, event data in a target area will be implicitly and/or explicitly collected, including the Wi-Fi transmissions and denoted positions of smartphones. A further step of data processing will be performed in the cloud server to generate event-related information, e.g. crowd density and velocity, which can be broadcast by our platform and shared by everyone in a community. Develops a crowdsourcing system to monitor public events Timely broadcasts valuable information to smartphone users. Secures event data (e.g. crowd density, population, moving speed, waiting time and location information) Crowd monitoring Public event surveillance Waiting time estimation of a queue Business promotion to users smartphones Funding Party Central Research Grant of The Hong Kong Polytechnic University for RGC General Research Fund Project Rated

72 New Technologies for Dynamic RFID Systems: Problems,, and Solutions 動態 RFID 對物體信息的高效收集和處理 Dr. XIAO Bin New technologies for dynamic RFID systems for efficient tag identification and information collection. The deployed RFID reader Our project aims to provide efficient and secure solutions to fundamental questions in the RFID research community by addressing the challenges from tag/reader dynamic property. Fundamental questions is concerned with tag cardinality (or population) estimation and ID collection. Unlike the previous work, our project will focus on dynamic tags, i.e. missing tags and new tags, and analyze their effects on tag information collection efficiency. A further step is to consider the use of mobile readers, which can assist in system management to relocate misplaced tagged items to their original positions, e.g. in a library or a supermarket. In addition, our project will design secure communication protocols in a malicious environment. Optimizes system performance to achieve time efficiency, energy efficiency and accuracy for tag identification and population estimation Designs secured solutions to explore attack detection and identification in anonymous RFID systems, enhancing privacy preservation Employs mobile readers to assist in information collection Provides efficient RFID data storage and query in logistics RFID technology used in the warehouse management Inventory management in logistics Library book management using the RFID technology Medicine and drugs management in hospitals and pharmacies Funding Parties National Natural Science Foundation of China RGC General Research Fund The RFID system with a reader and active tags 71

73 Wireless Indoor Localization System Design 智能手機用戶的無線室內定位系統 Dr. XIAO Bin New localization techniques through Wi-Fi or RFID to localize smartphones or RFID tags in indoor environments. Indoor localization has many significant applications in locating and tracking objects, e.g. goods in a warehouse, people in a building, etc., where traditional Global Positioning System localization technique cannot work properly. It has become increasingly important in the modern merchandise management for location query of misplaced or malfunctioned items. Previous approaches limit indoor localization by only investigating the constraints between tracked objects and reference nodes such as access points and anchors deployed at known positions. They have two limitations: low localization rate due to insufficient reference nodes (or wireless signal blocking) and low localization accuracy due to received irregular signals. Our project addresses these limitations and aims to accurately and economically localize indoor objects by new technologies, e.g., exploiting the Wi-Fi system to localize smartphones or the RFID system to localize RFID readers/tags. Our proposed indoor localization techniques can substantially improve the localization rate and accuracy through developing a real indoor localization system to track the movements of smartphone users. Develops an accurate indoor localization system to track the moving smartphones Provides location-based service to smartphone users Designs an RFID localization system to localize tagged objects, e.g. books, jewelleries, etc. High accuracy of errors within 2 3 meters A wireless localization system with mobile cars and wireless sensors Indoor localization to identify potential customers People and item localization, monitoring and tracking Business promotion to customers within a range Automatically identify suspicious behaviors of people within a sensitive region Funding Parties Central Research Grant of The Hong Kong Polytechnic University for RGC General Research Fund Project Rated 3.5 RGC Direct Allocation of The Hong Kong Polytechnic University for RGC General Research Fund Project Rated 3.5 A simulated environment to show the localization errors in our proposed approaches 72

74 Automatic Performance Tuning for Spatial Databases 空間數據庫自動化性能調整 Dr. YIU Man Lung System architecture Tuning the performance of a spatial index automatically. Since many location-based applications like inventory management and fleet management encounter massive workloads, it is crucial to effectively tune the database performance. The enormous workloads includes location updates issued by moving objects (e.g. containers) and spatial queries issued by users (e.g. warehouse staff) to locate objects in specified regions. Our project aims to develop automatic performance tuning techniques for spatial indexes, such as techniques for tuning the spatial index size, spatial indexes in commercial spatial databases, road network indexes as well as performance evaluation and prototype tuning tool. Improves the performance of spatial index Avoids manual tuning Fleet management Inventory management Funding Party RGC General Research Grant Object representation and operations 73

75 Enhancing Location-Based Services by Using Online Route Services 利用路線信息提升地理位置服務效能 Dr. YIU Man Lung System architecture Using online route APIs to improve the accuracy and efficiency of processing queries at location-based services. Location-based services (LBS) enable mobile users to query points-of-interest, e.g. restaurants, cafes, etc. for different features such as price, quality and variety. In addition, users require accurate query results with up-to-date travel times. The LBS can obtain live travel times of routes from online route Application Programming Interfaces (APIs) in order to offer accurate results, without the road traffic monitoring infrastructure. Our project aims to reduce the number of requests issued by the LBS significantly while preserving accurate query results. We are able to exploit recent routes requested from route APIs to answer queries accurately and design effective techniques to process queries efficiently. Provides accurate query results to end users Reduces the cost of LBS providers for using online route APIs Restaurant guide websites Shopping guide websites Short-term travel time approximation Funding Party Central Research Grant of The Hong Kong Polytechnic University for RGC General Research Grant Project Rated

76 A New Secured Retinal Imaging System for Computer Aided Non-intrusive Diabetic Care 新一代安全可靠非侵入式糖尿病監護的視網膜圖像系統 Professor YOU Jia Jane A novel medical imaging system that performs non-intrusive detection of Diabetic Retinopathy with a high-performance fundus camera and provides information security and privacy protection by effective lossless data embedding With the increasing population of diabetic patients worldwide, it is crucial to detect various diabetic complications timely, efficiently and conveniently for large-scale screening at low cost. However, the conventional blood glucose testing does not provide the progress status of diabetic chronic complications such as Diabetic Retinopathy (DR) for non-intrusive diabetic care. By developing new approaches to hierarchical retinal image analysis and lossless data embedding associated with a highperformance fundus camera, we have invented a computer aided detection (CAD) system to achieve non-intrusive DR screening and monitoring with privacy protection. The comprehensive performance assessment of our prototype demonstrates its feasibility for effective diabetic care with wide applications and excellent potentials. a special optical design for a new model of fundus camera for lighting normalization and position alignment a hierarchical approach to retinal image analysis by multiple feature extraction and coarse-to-fine lesion detection a wavelet based scheme for lossless data embedding to facilitate information security and privacy protection The structure of the new retinal imaging system An example of lossless data embedding: Embedding a face image into retinal image for privacy protection The new effective retinal image analysis software in conjunction with the high performance fundus camera has wide clinical applications not only by ophthalmologists for eye care but also by other medical specialists for general health services including healthcare and telemedicine The new algorithms for lossless data embedding will have significant contribution to information technology ranging from multimedia to information security Awards The 2nd place in SPIE Medical Imaging 2009 Retinopathy Online Challenge (2009) Special Prize and Gold Medal with Jury s Commendation at The 39th International Exhibition of Inventions of Geneva (2011) Patent No. & Country Methods of segmentation of color retinal images for DR (Diabetic Retinopathy) diagnosis (Patent Application No. 13/436,991, United States) Funding Parties Innovation Technology Fund RGC General Research Fund Illustration of hierarchical detection of retinopathy 75

77 Professor YOU Jia Jane Automated Truck Loading Monitoring for Smart Logistic Management and Road Safety Control 用於智能物流管理與道路安全的車載監控系統 A new platform that offers real-time truck loading monitoring with a portable multi-function tool kit and provides secured and optimized transportation information management by integrating sensing, mobile communication, task scheduling and Global Positioning System. We have developed an automated truck loading monitoring system by integrating sensing, mobile communication, Global Positioning System and logistic management for high performance. Unlike the existing systems with limited functions, this new vehicle-mounted dynamic monitoring tool-kit offers a wide range of services on a new platform including fast truck status checking, weight recording, loading data transmission/tracking/monitoring, position localization, logistic scheduling, user interface, security control and emergence management. This novel system has demonstrated its advantages and potentials to improve traffic control, enhance road safety, reduce maintenance costs on transportation, and increase logistic efficiency. Special Features and Advantages Low-cost and reliable sensors for load weight measurement On-board monitoring system for fast data recording, transmission, tracking and analysis Truck status monitoring and position tracking Optimized logistic scheduling Automated load monitoring to enhance payload management and road safety control Dynamic task scheduling and real-time condition checking for smart logistic management. The structure of the new load monitoring system Manufacturer PolyU s new product Truck Weight VEI group Country Hong Kong/China Canada Italy Sensor type Mechanical suspension Air suspension sensor & Air suspension sensor sensors mechanical suspension sensors Communication Wire and wireless Wire and wireless Wire Suitable trucks No limitations Trucks with air suspension Trucks with air suspension Accuracy 2% 5% 1% 2% 2% 3% Initialization less than 2 hours at least 24 hours at least 12 hours Power supply Truck battery AA battery Truck battery Water proof Yes Partially Yes Remote monitoring Yes Yes Yes Monitoring distance Free range < 500 feet < 500 feet Remote monitoring Bluetooth/3G/GPRS/ Zigbee Zigbee Zigbee technology Price <US$120 >US$500 >US$700 Sensor sample Performance comparison 76

78 Facial Beauty Analysis 人臉美學分析 Professor ZHANG Dapeng David Convex hull based facial beauty enhancement (A) Face images were deformed to its projection in the convex hull (B) Examples of faces before and after deformation Understanding facial beauty perception, building computational models of facial attractiveness, and enhancing automatic facial attractiveness. Attractiveness is an important attribute of a face that can be understood, predicted and manipulated. We analyze facial beauty using biometric techniques. Our research has three modules: evaluating and verifying the existing facial beauty theories based on data-driven approach; building computational models for automatic facial attractiveness prediction; and manipulating the attractiveness of a face with the guidance of the valid rules and the learned models. Extensive methods of image processing, machine learning, data mining, and statistics have been applied to promote our research. Several new findings and their potential applications have been obtained. Flow chart of automatic facial attractiveness prediction by PCANet, LPB, and AAM features and score level fusion A new hypothesis on facial beauty perception A computational model of facial attractiveness with a correlation of 0.9 between human rated scores and model predicted scores. A convex hull based facial beauty enhancement method Plastic surgery plans Cosmetic industry Photo retouching Entertainment Funding Party National Natural Science Foundation of China 77

79 Palmprint Recognition for Security 掌紋識別系統 Professor ZHANG Dapeng David Online palmprint identification system with low cost and high speed. Our low-cost, high speed and high reliability palmprint identification system, including palmprint image collection, palmprint recognition and output-input system, uses proprietary data collection device and advanced processing algorithm to measure stable line and texture feature on palms. It can identify a person with high accuracy and high speed, and it can outperform many high cost security systems with its relative low cost and high reliability. High accuracy: comparable to two highly accepted biometrics techniques (fingerprint and iris) Fast response: identify a user in seconds High user percentage: around 4% of the population could not use finger identification system. However palmprint identification system can be used by all people Liveness anti-spoofing detection Development of palmprint identification system Citizen-related applications (e.g. e-passports, border control, etc.) Forensic-related applications (e.g. motor vehicle security, monitoring, etc.) Employee-related applications (e.g. time, attendance, etc.) Award(s) Silver Medal at The Seoul International Invention Fair (2002) Special Gold Award and Gold Medal at the 14th National Invention Exhibition of China (2003) Hong Kong Award for Industry (2003 and 2009) Silver Medal at the 34th International Exhibition of Inventions, New Techniques and Products of Geneva (2006) The 1st Prize at 2011 Nature Science Award of palmprint identification system 78 Patent No. & Country Patent No.7,496,214, United States Patent No. 7,466,846, Unite States Patent No , Japan Patent No. HK , Hong Kong Patent No. ZL , China Patent No. ZL , China Patent No. ZL X, China Patent No. ZL X, China Patent No. ZL , China Funding Parties Commission of European Communities Competitive Earmarked Research Grant Innovation and Technology Fund National Natural Science Foundation of China National Natural Science Foundation of China Oversea Key Collaboration Fund Niche Area of The Hong Kong Polytechnic University

80 Tongue/Pulse/Odor Detection for Healthcare 健康診斷 : 舌象 / 脈象 / 氣味檢測 Professor ZHANG Dapeng David csdzhang@comp.polyu.edu.hk Detecting abnormal tongue/pulse/odor signals from human body for possible disease identification. Health is of great importance for humans. Diseases have external reflections on humans, such as changes in tongue, pulse and odor. By analyzing those changes in tongue, pulse and odor signals, we can diagnose diseases for health protection. Currently, over 13 kinds of diseases can be detected by our automated tongue analysis system. Novel pulse diagnosis system can provide more detailed information for diagnosis by combining pressure sensors with photoelectric ones. A breath analysis system, including a specialized breath analysis device, signal processing and pattern recognition algorithms, has been developed to aid the diagnosis or monitoring of some diseases, especially for diabetes detection. Combination of medicine, sensor technology and computers Non-invasive Fast Low-cost Convenient Portable Automated tongue analysis system A pulse analysis system A breath analysis system for diabetes detection Early diagnosis of diseases such as diabetes and heart disease Non-invasive monitoring of diseases such as diabetes and chronic renal disease Research on gas detection, pulse detection, medical diagnosis and biometrics; education on Traditional Chines Medicine Patent No. & Country Patent No. 8,355,543 B2, United States Patent No. ZL , China Patent No. ZL X, China Patent No.ZL , China Patent No. ZL , China Patent No. ZL , China Patent No. ZL , China Patent No , China Patent No. ZL , China Patent No. ZL , China Patent No. ZL , China Patent No , China Patent Application No. CN , China Patent Application No. CN X, China Patent Application No. CN , China Patent Application No.CN Funding Parties National Science Foundation of China ( 重大項目 ) Hong Kong Hospital Authority 深圳市戰略新興產業發展專項資金 79

81 Fast Compressive Tracking 基於視頻的高效目標跟踪技術 Dr. ZHANG Lei A highly efficient and robust tracking algorithm. It is a challenging task to develop effective and efficient appearance models for robust object tracking due to factors such as pose variations, illumination changes, occlusions and motion blur. We have proposed a simple yet effective and efficient tracking algorithm with an appearance model based on random projections that preserve the structure of the image as well as feature and space of the objects. A very sparse measurement matrix has been adopted to efficiently extract the features for the appearance model. We have compressed samples of foreground targets and the backgrounds using the same sparse measurement matrix. The tracking task has been formulated as a binary classification through a naïve Bayesian classifier with online update in the compressed domain. The proposed compressive tracking algorithm can run in real time and perform favorably against state-of-the-art algorithms on challenging sequences in terms of efficiency, accuracy and robustness. Fast-running algorithm (up to 150 frames/second in MATLAB) Robust algorithm to various target variations Easy to be extended and combined with other algorithms Examples of target tracking High speed target tracking Embedded into video surveillance systems for robust target tracking Real-time robust object tracking in devices with limited computational resources such as tablets, mobile phones, etc. Target localization in video editing Funding Party RGC General Research Fund 80 Framework of our algorithm

82 Image and Video Resolution and Quality Enhancement 圖像與視頻分辨率和品質增強系統 Dr. ZHANG Lei We make your images or videos with higher quality and resolution. Due to the limitation of low-end consumer electronics imaging sensors, the captured images and videos often result in low resolution and quality. Furthermore, the resolution and visual quality of Internet Protocol TV (IPTV) videos are unable to meet the users requirements due to the network bandwidth problem. Image and video super-resolution is one promising technique to enhance the resolution and visual quality of images and videos. However, few products are available in the current market for real-time resolution and quality enhancement. In this project, we have developed a set of software and algorithms to produce high-performance images and videos as well as enhance their quality in real time. An example of super-resolution Enhances the resolution of images and videos while keeping the edges and textures of the image and video contents Produces real-time IPTV video stream resolution and quality enhancement under low bit rate Improves the resolution and quality of pictures and videos captured by video cameras, camera phones, etc. Enhances the resolution and quality of IPTV in real time Enhances the resolution and quality of videos downloaded from Internet Enlarges and sharpens the pictures for large-sized poster printing A low resolution image Award Silver Medal at The 62nd International Trade Fair "Idea-Inventions-New Products" (iena) (2010) Patent No. & Country PCA based noise reduction method for color filter array (CFA) data of digital video camera (Patent No. 8,345,971, United States) A resolution-enhanced image by our innovative technology Funding Parties Innovation and Technology Fund RGC General ResearchFund News in Sing Tao Daily 81

83 Low Cost and High Performance 3D Printing System 低成本高性能 3D 打印系统 Dr. ZHANG Lei The developers (Dr. ZHANG Lei [left] and Dr. CHENG Junting [right]) We print what your imagine. Example artwork produced by our 3D printer 3D printing integrates the latest achievements of Numerical Control, CAD/CAM, Computer Science, Laser, Materials, etc. It has high interchangeability, can be easily replicated, and can manufacture prototypes with complex surface, etc. However, the current 3D printing systems in the market are not only expensive but also limited in functionality. In this project, we have developed a low-cost high-performance desktop 3D printing system, which has much lower cost, higher printing precision, faster printing speed and larger printing size than the existing systems. In addition, our 3D printing system can be widely adopted for office, family and industrial use. Low cost: the cost of our system (<HK$5,000) is much lower than the existing commercial systems High precision: it has higher printing precision than many more expensive systems High printing speed of up to 800 mm/s Large printing size of π 8 cm 2 22 cm The outlook of the 3D printer 3D prototype printing of artwork, toys, etc. 3D customized souvenir printing 3D family photo printing 3D tourism industry, e.g. printing 3D face or portraits for tourists with the background as the landscape Funding Party Lab Fund by Department of Computing of The Hong Kong Polytechnic University 82

84 Personal Authentication Using Finger-Knuckle-Print 基於指節紋的身份認證系統 Dr. ZHANG Lei An accurate and user-friendly biometrics technique. Recognizing the identity of a person is a critical issue in various applications, traditional methods can only identify persons according to what they have (e.g. key) and what they know (e.g. password). Yet both are easily lost or forgotten. Biometric techniques automatically authenticate persons through human physical or behavioral traits. However, limitations are found from the biometrics systems available in the current market. For example, fingerprints are restricted by specific population, face identification has low accuracy and iris checking might cause stimulation to human eyes. This low-cost and highly efficient Finger-Knuckle-Print Identification System, including image collection, recognition and outputinput modules, employs advanced pattern recognition algorithms to measure stable lines and texture features on the finger knuckle surface. It can quickly and accurately identify a person s identity, and it outperforms many other expensive security systems with lower cost. High accuracy: comparable to other biometrics techniques Fast speed: can identify a user in 1 to 2 seconds High user friendliness: easy to use, no stigma of criminal investigation High anti-counterfeit capability: difficult to forge Highly reliable: suitable for both indoor and outdoor environment The outlook of our Finger-Knuckle-Print recognition system (embedded system) A captured Finger-Knuckle-Print image Attendance system for various organizations Physical access control for organizations, buildings, etc. Credit cards Used in hotels to enhance their security on specified guest rooms and other facilities Citizen ID cards Military forbidden, social security system and ATM Award Best Paper Award (Honorable Mention) awarded by the journal Pattern Recognition for the paper Lin Zhang et al., Online finger-knuckle-print verification for personal authentication, Pattern Recognition 43 (7) , 2010 (2013) Patent No. & Country Method and system for identifying a person using their finger-joint print (Patent No. US8,208,692 B2, United States) 指關節紋識別個人身份的方法和系統 ( 中國發明專利號 : ) Funding Parties Edward Sai Kim Hotung Fund Internal Competitive Research Grant funded by The Hong Kong Polytechnic University 83