PhD Project Proposals: Info-Communications

Transcription

1 Andy Khong Data Analytics for Learning Science The aim of this research is to develop data analytics to understand human learning behaviour. This project involves data mining and machine learning which is applied on understanding human behaviour as they go through online courses. This interdisciplinary research will incorporate learning science and learning pedagogy. Andy Khong Acoustic signal processing and acoustic enhancement technologies The aim of this research is to develop algorithms to achieve audio enhancement in a noisy environment. This involves the development of array signal processing algorithms, which can applied to microphone array in a noisy environment. Potential applications include surveillance and meeting scenarios where the quality of sound is important.

2 Arokiaswami Alphones Metamaterials and Peridoic structures Recently leaky wave phenomena from left handed wave guides are very interesting research topic, especially for their composite forms of left and right handed transmission line where wave can be scanned from backward to forward directions continuously through zero order propagation constant by changing frequency. They are expected small leaky wave antenna for communication use. Few years back we have proposed double periodic left handed transmission line where capacitance or inductance is periodically loaded in the line. Realizing compact antennas and filters using metamaterials concept would be attempted in this work. Arokiaswami Alphones Metamaterial assisted Wireless Power Transfer System WIRELESS POWER TRANSFER (WPT) systems have been extensively investigated in the last few years. WPT technology can be used in a wide range of industrial applications such as biomedical implants, robotics, consumer electronics and electric vehicle (EV) charging to eliminate charging hazards and drawbacks related to cables. In this project, an analytical method to calculate the scattering parameters of a wireless power transmission link composed of electrically small single loop resonators will be attempted. The method has to consider all the different couplings in the structure. First, the approach to be used to find the S-parameters for links composed of circular and rectangular resonators. The model is then to be used to find the optimal topology for a given transmission distance. Validation of the model has to be done by comparing its results with experimental measurements.

3 Proposed Research Project (3): Arokiaswami Alphones Perturbation Analysis on Grating Coupler Optical integrated circuits consist of multiplexers,input, output couplers. This project will theoretically investigate with rigorous analysis on periodically modulated grating structures in optical domain. Chau Lap Pui Human motion retrieval system based on human skeleton extraction As motion capture device becomes more popular, the motion data sets are becoming larger. So efficient retrieval method for the human motion database is very important. This project is to develop an efficient retrieval method that requires efficient indexing scheme and robust similarity measure to represent motion data.

4 Chau Lap Pui Contrast Enhanced Vision for Deepwater Monitoring System Current approaches to image processing for contrast enhancement rely on standard models of signal-to-noise ratio (SNR). Novel optical characterization with associated signal processing can enable dramatic increase in image and video sharpness and clarity. The proposed research can apply to vision based operation under deepwater environment. This proposal will describe a feasible method to enhance deepwater optical image and video. Deepwater vehicles namely Remotely Operated Vehicles (ROV) generally comes with optical sensors for their capability of remote operation. Therefore, deepwater vision is an important issue in maritime application. Different from the common video, deepwater video suffer from poor visibility due to the scattering and light distortion. Owing to the challenging deepwater conditions and the light dissemination, conventional image and video processing methods cannot be applied directly in deepwater video. The proposed research intends to alleviate the drawbacks of the low light saturation to obtain a clear vision. Proposed Research Project (3): Chau Lap Pui elpchau@ntu.edu.sg Feature Extraction and Representation for Human Action Recognition In this project, we would like to propose a new framework for the real-time realization of human action recognition. New feature descriptors are developed that are tolerant for illumination change which can be run in various configurations on a platform suitable for use as an embedded system. This proposed framework will also be extended to support multiple camera in order to improve the accuracy.

5 Jiang Xudong Machine learning based feature extraction from image and audio signal Feature extraction takes a key role in the image and audio signal recognition. Yet there lacks effective techniques to extract discriminative and reliable features from image and audio signal because we have very limited expert knowledge about these features. Machine learning from database is a solution. However, how to make the learning results from a specific database general to the whole population is a challenging research issue worth to be investigated. Project Outline: Study the traditional techniques in feature extraction from image and audio signal; Study various machine learning approaches in feature extraction and dimensionality reduction; Explore the limitations and problems of the stateof-the-art technologies in these two areas; Develop robust new approaches to solve generalization problem of the machine learning in feature extraction. Evaluate and apply the developed approaches on the image object detection, recognition and speech recognition. Jiang Xudong exdjiang@ntu.edu.sg High-dimensional, Incomplete And Unbalanced Data Analysis In many real applications such as visual object detection and recognition, bioinformatics, and data mining, high data dimensionality imposes great burdens on the robust and accurate recognition due to insufficient knowledge about the data population and limited number of training samples. Incomplete data and/or unbalanced data make the knowledge discovery from the data even more difficult. They often cause biased estimation. While there are some techniques to handle the issues of high dimensionality of the data, problems of incomplete data and/or unbalanced data are new, which are very challenging and not well studied. This project aims at solving these challenging problems by generate novel techniques in machine learning, statistical estimation and pattern recognition.

6 Proposed Research Project (3): Jiang Xudong Sparse Representation and Dictionary Learning for Computer Vision Sparse signal representation has proven to be an extremely powerful tool for acquiring, representing, and compressing high-dimensional signals. This success is mainly due to the fact that important classes of signals such as audio and images have naturally sparse representations with respect to fixed bases. In computer vision, we are often more interested in the content or semantics of an image rather than a compact, high-fidelity representation. Representing a signal involves the choice of a dictionary, which is the set of elementary signals used to decompose the signal. When the dictionary forms a basis, every signal is uniquely represented as the linear combination of the dictionary atoms. This project aims at solving difficult computer vision problems through the sparse representation of images based on proper learnt dictionary.

7 Law Choi Look Ultra low power wireless communication devices for E- sensing in dense cluttered environments Indoor environments and underground environments are examples of dense cluttered environments. These environments are known to be hostile to wireless communication signals due to multipath signals. Current low power wireless communication technologies such as Zigbee and Bluetooth are highly susceptible to multipath signals (suffers from flat fading), resulting in degraded performances. Wide bandwidth signaling (such as OFDM) is currently used to overcome the impairments caused by multipath signals. However, complex transceiver circuits are used resulting in increased power consumption. Ultra-wideband impulse radio (UWB-IR) is an innovative signaling method whereby the carrier frequency is turned on for a few cycles in a data symbol. This short pulse duration results in multipath signal arrivals which do not interfere with the direct path transmission. Precise ranging based on time of arrival of the direct path signal can also be extracted. The objective of this proposed project is to investigate hardware circuits for energy efficient transmission of environmental data (e.g temperature, light intensity, etc) from numerous spatially distributed sources to a central receiver using UWB-IR signals. The investigations include novel circuits for energy efficient generation of UWB-IR electromagnetic waves with superimposed data modulation and energy harvesting circuits for powering up the data source. Novel antenna array techniques exploiting the UWB-IR multipath channel characteristics will be investigated to improve the receiver sensitivity.

8 Law Choi Look Energy efficient large scale wireless sensor network for E- sensing in dense cluttered environments Indoor environments and underground environments are examples of dense cluttered environments. These environments are known to be hostile to wireless communication signals due to multipath signals. Current low power wireless communication technologies such as Zigbee and Bluetooth are highly susceptible to multipath signals (suffers from flat fading), resulting in degraded performances. Wide bandwidth signaling (such as OFDM) is currently used to overcome the impairments caused by multipath signals. However, complex transceiver circuits are used resulting in increased power consumption. Ultra-wideband impulse radio with its large signal bandwidth and simple transceiver circuits is well suited for ultra low power wireless communication E- sensing in dense cluttered environments. The objective of this proposed project is to investigate medium access control and network protocols for energy efficient large scale wireless sensor networks with simultaneous nodes localization capability. These wireless sensor networks are expected to work in dense cluttered environments for E-sensing applications such as environmental monitoring, industrial monitoring and target tracking. IR-UWB physical layer will be used in the study. Channel parameters sensed from the IR-UWB signals will be used to optimize time slot allocation in medium access control, multi-hop routing and nodes localization. The study is expected to yield optimized sensor nodes clustering, time slot allocation, delay in sensed information transfer to cluster head and tradeoff with energy usage.

9 Proposed Research Project : Li Kwok Hung ekhli@ntu.edu.sg Optimizing Traffic Management of Heterogeneous Wireless Networks A heterogeneous wireless network consists of devices using various radio access technology. One popular scenario is the inclusion of pico-cell and femto-cell networks within the traditional macro-cell networks. This type of networks is expected to be deployed in the next decade. The mixture of different devices leads to new challenges. There are many interesting issues still to be solved in heterogeneous wireless network such as interference management, dynamic resource allocation and green wireless transmission. The concept of heterogeneous wireless networks also opens up new research possibilities such as efficient and optimized traffic management for individual users to increase overall network performance, and application in the new "Internet of Things" paradigm. The research will focus on solving problems related to heterogeneous wireless networks and also introducing innovations within this new framework to improve the technology overall.

10 Ma Maode Security Enhancements for the M2M Communication in the Internet of Things The Internet of Things (IoT) will consist of billions of digital devices, people, services and other physical objects having the potential to seamlessly connect, interact and exchange information about themselves and their environment. It helps to optimize operations, reduce costs, boost productivity and improve lives. IoT represents the next evolution of the Internet, taking a huge leap in its ability to gather, analyze, and distribute data that we can turn into information, knowledge, and, ultimately, wisdom. Thus, in this project, a security scheme will be constructed to improve the security of IoT in the layer of 6LoWPAN with the consideration of large scale structure and resource constraint. Ma Maode emdma@ntu.edu.sg Enhancements of Cyber Security for Smart Grid Systems The smart grid is a promising power delivery infrastructure integrated with communication and information technologies. It could generate enormous economic, environmental and social benefits through the bi-directional communication and electricity flow, as well as the integration of vast distributed energy resources. In order to address such threats, existing security standards and technologies should be closely examined and investigated to ensure the feasibility in the smart grid systems. In addition, new communication protocol or security schemes will be designed supported by popular log ic theorem to prove the correctness, and be tested through simulation to prove the robustness and scalability.

11 Soong Boon Hee Network Optimization for interference Management for Heterogenous Cognitive Networks Over recent years, there has been a systematic increase of the requirements for the wireless resources requested by the mobile users. This has led to the increasing deployment of dedicated small-cells, where the considered small geographical area is served by low-range access points (or small base stations) creating a heterogeneous deployment hese deployments allow significant capacity enhancements and better coverage conditions that can be particularly beneficial for enhancing the bit rates of the users (spectral efficiency) that are located at the edges of the macrocells. Such advanced topologies composed by multiple cell sizes and involving the combination of different radio access technologies are usually known as heterogeneous networks or simply as HetNets. While some of the small-cells devices are deployed and installed by the operators e.g. in the street infrastructure, others (e.g. femtocells) are installed directly by the end-users offering portability and exclusive access to the spectrum within the user premises. On the other hand significant reduction of the cell radius results immediately in transmit power decrease as well as in creating of high potential for smart and effective resource and interference management to increase the system capacity. Moreover, the application of various wireless technologies used by the end users for spectrum access provides to the cellular network operators attractive opportunity for efficient traffic offloading to e.g. WiFi networks. Several research problems for this research work involves to for these heterogeous networks i.e. cognitive radio technologies for small-cells, Interference management and mitigation issue and Game theoretic algorithms management of dynamic spectrum allocation. Students should have a good background in research and programming.

12 Soong Boon Hee Design of Ambient Wireless Sensor Networks Recent advances in hardware and wireless technology bring us the phenonmenal reality of accesses information anywhere, anytime through the collection of wireless sensor nodes (WSNs), RFID tags that forms the collection of Internet of Things. These networks are form by ad hoc sensor nodes deployed in an fashion in our environment for physical sensing and monitoring purposes to achieve a functional objectives. There are numerous applications for these smart sensor enabled devices with RF enable wireless communication capabilities that would help connect a large number of measurement points at low cost without the need for cabling so that conducting measurement e.g. in unsafe or harsh environment is of special interest. The research would involve the design and development of wireless sensor interfaces that would be helpful for the following technologies e.g. wireless USB, health monitoring, location tracking and surveillance. This would be extended to multiple sensors that enable fusion of information and processing. The scope of research include the following (1) Study Ambient Wireless Sensor Systems (2) Design of DSP algorithms in embedded systems, data management and storage, diagnostic decision making. Students should have a good background in research and programming.

13 Teh Kah Chan Studies robust and energy-efficient cooperative communication systems over fading channels In this project, student will first conduct literature survey on existing cooperative communication systems. Some existing results will also be reproduced and compared. The effect of fading channel will be taken into consideration. Following that, new robust and energy-efficient algorithms will be proposed to improve the system performance. The corresponding system performance and complexity will be compared with some existing techniques. Teh Kah Chan Development and study of novel frequency-hopped multipleaccess receiver Recently, frequency-hopped spread-spectrum systems have been widely used in commercial and military applications. In this project, the student will first conduct literature survey on existing frequency-hopped multiple-access systems. Following that, the student will study and develop a novel receiver structure for frequency-hopped multiple-access systems over fading channels. Maximum-likelihood receiver will be proposed and Matlab simulation will be conducted to study the bit-error rate performance of the proposed receiver structure. Xiao Gaoxi egxxiao@ntu.edu.sg Resilience of complex systems: evaluation and enhancement There is no commonly accepted definition of what is system resilience though more and more people are talking about it. This project aims to develop a set of methods for measuring system resilience and predicting the system behaviors under disasters, as well as enhancing system resilience in average and worst cases respectively.

14 Xiao Gaoxi Coevolution of opinions and connections in complex social networks In this project, we shall study on the complex dynamics of the co-evolution of opinions and their underlying social networks. Benefiting from the latest developments in complex systems and multi-agent systems, we shall propose new models and novel analyses that will provide useful insights into the interesting co-evolution of social opinions and social networks. Proposed Research Project (3): Xiao Gaoxi Control and controllability of complex networks and applications In this project, we shall study on the control and controllability problems for complex systems and their applications. We shall investigate the possibility of developing new distributed, fault-tolerant methods to achieve effective, robust and scalable control of various complex systems. Yap Kim Hui Mobile Visual Object Search The objective of mobile visual search is to recognize the objects of the captured images, and then retrieve relevant information of the recognized objects. An example of such applications is mobile consumer product recognition and recommendation service where the user can capture the image of a consumer product (e.g. book/cd/dvd cover, logo, advertisements, etc.) and then use the system to recognize it and retrieve relevant information such as the brand name, the price comparison and recommendation, etc. The project will study key issues including visual content analysis, large-scale visual search and recognition, and user-system interaction.

15 Yap Kim Hui Mobile Media Content and Context Analysis The objective of this project is to study and develop content and context analysis for mobile media applications. In particular, it aims to develop and transform a mobile device into a portable information terminal that renders various situation-aware mobile services, e.g. mobile landmark recognition, augmented-reality city guide, and consumer product recognition. The project will investigate key issues including content and context analysis, efficient visual search and recognition, and user-system interaction. Proposed Research Project (3): Yap Kim Hui ekhyap@ntu.edu.sg Mobile Media Visual Object Search The growing usage of mobile devices has led to proliferation of many mobile applications. Amongst them, mobile media visual object search is an emerging application that is receiving growing attention recently. The objective of mobile visual search is to recognize the objects of the captured images, and then retrieve relevant information of the recognized objects. An example of such applications is mobile consumer product recognition and recommendation service where the user can capture the image of a consumer product (e.g. book/cd/dvd cover, logo, advertisements, etc.) and then use the system to recognize it and retrieve relevant information such as the brand name, the price comparison and recommendation, the shops nearby selling the product, etc. The project will study key issues including visual content analysis, large-scale visual search and recognition, and usersystem interaction. An example of the developed system similar to this project can be viewed at Google Goggle Lab video

16 Proposed Research Project (4): Yap Kim Hui Media Content and Context Analysis and Applications The objective of this project is to study and develop content and context analysis for mobile media applications. In particular, it aims to develop and transform a mobile device into a portable information terminal that renders various situation-aware mobile services, e.g. mobile landmark recognition, augmented-reality city guide, and consumer product recognition. The project will investigate key issues including content and context analysis, efficient visual search and recognition, and user-system interaction. An example of the developed system similar to this project can be viewed at Google Goggle Lab video The project will also aim to address the challenges of mobile environments including low/moderate processing power of mobile devices, fast response-time requirement of users, and limited battery life of mobile phones. Yuan Junsong jsyuan@ntu.edu.sg Understanding human interaction in RGB-D videos Understanding human behaviors in real environments is one of the most pressing challenges to human-centered computing. Because nobody exists in a vacuum, the constant interactions among humans and those between humans and surrounding objects not only capture what we do, but also reveal our social sphere and who we are. Therefore, the ability to instantly detect and recognize these interactions is crucial to the creation of novel computing platforms that facilitate productive human-computer integration. This project targets at recognizing human-human interactions, e.g., hand waving, using RGB-D sensor, e.g., Microsoft XBox Kinect. The student will have opportunity to collaborate with leading researchers at Microsoft Research Redmond, USA.

17 Yuan Junsong Large-Scale Video Analytics and Search This project will work on large-scale video content analysis and search, such as video anomaly detection and human activity detection. Wei Lei Multimaterial fibres for high power mid-wave and longwave infrared transmission The broad objective of the proposed research is to fundamentally address the needs and challenges for developing low-loss and flexible specialty optical fibers in mid-wave and long-wave infrared region. By strongly confining light to a hollow core and suppressing materials absorption and nonlinearities, the resulting fibers can deliver high-peak power laser pulses in the wavelength range of 2-15 μm for various applications in medical treatment, industry processing, space science, sensing. Wei Lei wei.lei@ntu.edu.sg Structured functional spheres generated by an in-fibre fluid instability (Electronics and Info-Communications) The scope of this project is to study a new approach to realize in-fiber capillary instabilities and to harness this effect to fabricate functionalized micro-/nano-structures. The key objectives of the project include: (1) To fundamentally study the in-fiber capillary instabilities using simulation tools. (2) To design and fabricate fibers comprising different materials with different viscosities. (3) To build up the experimental setup to achieve in-fiber capillary instabilities. (4) To apply the produced functionalized structures for chemical sensing and biomedical applications.

18 Proposed Research Project : Liu Zheng z.liu@ntu.edu.sg Atomic 2D nanomaterials based wearable/flexible sensing electronics for personal health-monitoring system (Electronics and Info-Communications) Monitoring human physiological signals is considered to be an effective way for disease diagnosis and health assessment. Conventional hospital-centered healthcare sensing devices including infrared based photo-electric devices and rigid multi-electrode pressure sensors have been employed for physiological signals detection, however the use is yet rather limited because of their poor portability and wearability. Currently, 2D nanomaterials, including graphene, chemical modified graphene, and MoS2, have been widely investigated and shown unprecedented performance in flexible sensing electronics with ultrasensitivity, fast-response time and low power consumption etc. The good performance makes the flexible sensing electronics have the capability of monitoring human physiological signals such as heart rate, wrist pulse, muscle movement, and pressure from sole of foots. Therefore, the 2D nanomaterials were considered to be the most promising materials for the next generation of wearable/portable healthmonitoring systems. To improve the comfortability, convenience, and security of health-monitoring in clinical diagnoses and daily life, we proposed a newtype of cost-effective portable personal healthmonitoring system, which make use of the flexible and wearable sensing electronics to address the human health conditions. In this project, the atomic twodimensional (2D) layer nanomaterials were employed as electrode and channel materials to construct flexible sensing electronics.