Bio-Metric Authentication of an User using Hand Gesture Recognition

Transcription

1 Bio-Metric Authentication of an User using Hand Gesture Recognition Parashuram Baraki Doctoral Student, Jain University, Bangalore & Associate Professor, CSE Department, S. K. S. V. M. Agadi College of Engineering & Technology, Lakshmeshwar, Karnataka, India. Dr. V. Ramaswamy Professor and Head, CSE Department, S R C, Kumbakonam, SASTRA University, Tamil Nadu, India. Abstract This paper presents an authentication of a user based on gesture recognition algorithm using accelerometer. It is very challenging task to recognize user action and reaction corresponding to that in real time by designing a dynamic gesture interface system. Physical movements of face, fingers, hands or body with respect to environment action is referred as Gesture. For applications like sign language, virtual reality and remote control gestural interfaces are developed using gesture recognition and motion tracking techniques. In this paper, we are designing a hardware module consisting of a triaxial accelerometer and microcontroller transmission module to sense the hand gesture and collect the same. User can use this hardware module to draw any kind of gesture. Personal computer receives the hand motions transmitted by the transmitter that is connected to accelerometer. Using the proposed algorithm we are able to authenticate the user effectively. For continuous reaction gesture is recognized and estimates variations in gestures in our module. Our work has been found to provide better accuracy and validation result compared to existing works. Keywords: Accelerometer, gestures, hand written recognition, gesture algorithm and Raspberry Pi board, Pin Micro Controller. Introduction The development witnessed in electronic circuits during the last decade has resulted in smaller, lighter and smart consumer products of electronics like hand held computers and smart phones. As a result our day to day life has become very convenient. To sense the activities of human embedded system devices with inertial sensors are developed. The device is capable of capturing their motion trajectory information from accelerations which can be very useful for recognizing gestures. For general motion sensing inertial sensors have some advantages, without any external reference and operating limitation conditions inertial sensors can be operated. To generate different motion trajectories motion trajectories have different styles and speeds, which makes it difficult for common users. Several researchers have attempted to increase the accuracy of gesture recognition systems. A miniature MEMS accelerometer based recognition system can be used to acknowledge different kinds of gestures drawn by user. MEM stands for micro electro mechanical system. It is a type of accelerometer which makes use of micro-fabrication technology. Silicon is the prominence part of MEMS. Low usage of material consumption can be obtained by usage of MEMS. By reduction of size and mass applicability is increased. There can be direct interaction between MEMS on silicon and electronics, instead of connecting serial of external components connected or soldered on board. These are referred as smart integrated MEMS and include filtering, communication interfacing, data storage, data acquisition and networking. Not only by making things smaller MEMS will also makes them better. Device that measures physical accelerations are known as accelerometers.. Light, force, pressure etc. are examples for physical parameters. Related Work Accelerometer or gyroscope based and vision based are the two main existing types of gesture recognition methods. Vision based method have some disadvantages they are large data collection and processing, slower dynamic response and ambient optical noise. By using inertial measurement unit we implement recognition system which is based on MEMS accelerometer sensors. It will be burden in using gyroscope for inertial measurement [2]. Hence our system is based on MEMS accelerometers. To improve recognition accuracy one has to compensate errors of inertial sensors, to achieve this many researchers have focused on error compensation. To track trajectories in 3-D space with the help of gyroscopes and accelerometers Yang et al. [3] proposed pen type input device. In acquiring accurate reconstructed trajectory acceleration errors will be there to avoid that or to compensate that zero velocity compensation based efficient acceleration error compensation algorithm has been developed. A developed Kalman filter with magnetometers (micro inertial measurement unit (μimu) with magnetometers), proposed by 4118

2 Luo et al. [9], was utilized to remunerate the introduction of the proposed advanced written work instrument. Accurate reconstruction of motion trajectories of instrument can be obtained by estimating the orientation of the instrument. Dong et al. [4] have proposed an optical following alignment technique in view of optical tracking system (OTS) to adjust angular velocities and 3-D accelerations. The OTS has been developed with the subsequent 2 goals: 1) to get accelerations of the projected ubiquitous digital writing instrument (UDWI) by calibrating 2-D trajectories and 2) to get the correct angles by mistreatment multiple camera activity. However, so as to acknowledge or reconstruct motion trajectories accurately sensors like gyroscopes or magnetometers square measure used for getting precise orientation. This will increase not solely value for motion flight recognition systems however conjointly the machine burden. In this paper, a conveyable device has been developed with a hand gesture recognition algorithmic rule. The moveable device consists of a tri axial accelerometer and a microcontroller module. The acceleration signals measured from the tri axial accelerometer square measure transmitted to a pc. Users will utilize this portal device to draw gestures at traditional speed. The measured acceleration signals of those motions may be recognized by the hand gesture recognition algorithmic program. The popularity procedure consists of AN acceleration acquisition, signal preprocessing, feature generation, feature choice and have extraction. The acceleration signals of hand motion are measured by the moveable device. The signal preprocessing procedure consists of activity, a moving average filter, a high-pass filter and normalization. First, the accelerations are tag to get rid of drift errors and offsets from the raw signals. These 2 filters are applied to get rid of high frequency noise and attraction acceleration from the information. The options of the preprocessed acceleration signals of every axis include mean, correlation among axis and mean absolute deviation (MAD). Before classifying the hand motion trajectories, we tend to perform feature choice and extraction procedures. In general, feature choice aims at choosing a set of size m from an inventive set of d options. The criterion of kernel-based category separability (KBCS) with best individual N (BIN) is to pick important options from the first options (picking up some vital options from the first set of features) which of linear discriminate analysis (LDA) is to scale back the dimension of the feature area with a higher recognition performance (reducing the scale of m). The target of the feature choice and have extraction strategies isn't solely to cut back the burden of process load however additionally to extend the accuracy of classification. The reduced options square measure used as inputs to the classifiers. The contributions of this paper embody the following: 1) the event of a portable device with the hardware module that helps user to draw gesture and 2) an efficient hand gesture recognition rule. In different words the projected methodology is anticipated to with efficiency choose important options from the time and frequency domains of acceleration signals and project the feature area into a smaller feature dimension for motion recognition with high recognition accuracy. Hardware Design The Raspberry Pi2 board is a small microcontroller with sufficient computing power to execute any kind of experimental programming. It provides all the necessary hardware along with Operating system like Raspbian Linux OS. It is easy to setup and use. Raspberry contains Broadcom processor BCM2836 (quad core ARMv7) with 1GB RAM. BCM2836 is a System on Chip (SoC) which means SoC combines the processor, graphics processing unit (GPU) and memory are combined in to a single chip. Other hardware features are USB, Ethernet, A/V, HDMI, micro SD, micro USB connectors and 40-pin GPIO. The fig1 shows all the components of a Raspberry Pi2 board. The power requirement of Raspberry Pi is more compared to other microcontrollers. Raspberry Pi has Quad-core ARMv7 processor means it draws higher current even if Pi 2 Model B is allowed to run idle (no HDMI, graphics, Ethernet or WiFi just console cable) will draw 200mA. With WiFi another 170mA is added. Ethernet adds about 40mA. Performing with heaviest computational tasks will add about mA current and WiFi dongle one will require at least 650mA at 5V. To start a Raspberry Pi 2 following basic components are required. SD Card: The basic requirement is 8GB SD card. This SD card acts as secondary memory (like Hard disk in a PC). Several Operating systems are available or supported by Raspberry Pi. Most widely used is NOOBS package which contains Raspbian Linux OS. Display and connectivity cables: Any Monitor with HDMI or DVI interface should work as monitor for Raspberry Pi board. LAN Access: Standard Ethernet cable can be used for LAN or Internet connection. Keyboard and Mouse: The USB slots can be used to connect keyboard and mouse. Power supply: Raspberry Pi can be powered by using a 5V micro USB power supply. Care should be taken with power supply as its output should be at least 5V. The typical Raspberry connections are shown in the diagram given below. 4119

3 Figure 2 shows the given below block diagram of portable device in which a user can draw his/her gesture through accelerometer. That gesture will be recorded in the micro controller and the same gesture will be displayed on the monitor as input during recording time and displayed as output when a match occurs with available gestures in the database. The connection and description of each part of the portable device are also shown in figure. Figure 2: The block diagram of the portable device Hand Gesture Recognition Approach When the sensing system is switched on, the accelerations in 3 perpendicular directions area unit detected by the sensors and gesture drawn by the user is transmitted to a pc. The gesture motion knowledge then goes through a segmentation program that automatically identifies the beginning and finish of every gesture so solely the data between these terminal points are going to be processed to extract options. Subsequently, the processed knowledge area unit recognized by a comparison program which can confirm the gestures given. The work flow of this method is shown in Fig 3. Data Acquisition: For sensing system reliable hand gesture data is collected, the sensing devices should be held properly during the whole data collection process. Gesture Segmentation: This includes data preprocessing during which raw data received from the sensors are preprocessed. the aim of segmentation algorithm is to seek out the terminal points of every gesture in a very information set of gesture sequence. The algorithm checks numerous conditions of all points and picks out the most probably data points because the gesture termination points. Figure 1: RaspberryPi2 Components and GPIO Pins General Classification: Gestures may be of 1 dimension, 2 dimensions or 3 dimensions. The gestures drawn by the user are classified and identified. They are then compared with gestures stored in the database and the result is displayed. The microcontroller collects the acceleration signals of hand motions that are generated by the accelerometer. Attributable to slight tremble movement of hand it's doable that specific amount of noise is generated. Signal preprocessing consists of activity, a moving average filter, a high-pass filter and normalization. Firstly, the accelerations are mark to get rid of drift errors and offsets from the raw signals. The second step of signal preprocessing is to use a moving average filter to cut back the high-frequency noise of the mark accelerations of gesture. The characteristics of different hand movement signals can be obtained by extracting features from the preprocessed x-axis, y-axis, and z-axis signals. Different features are extracted from the tri-axial acceleration signals. 4120

4 Once the registration phase is completed, whenever user wants to log into the system, the system verify whether the user is an authenticated user or not by using verification phase. In the verification phase, when user draws the gesture then that gesture will be compared with all other gestures which are stored in the database. If the drawn gesture matches with one of the available gestures in the database within a given threshold value, then the user will be considered as an authenticated user. Otherwise, the user is declared to be not a valid user. The following figure shows the result of authentication. There are two windows one showing the registered gesture and other showing current matching gesture. If both gestures are similar then the authentication is successful and the name of the user is displayed. If there is a mismatch between two gestures, the user is declared as unauthenticated. We have conducted experiment with five users. Three users have been declared as authenticated users and two unauthorized users. Figure 3: Workflow of gesture recognition system using Accelerometer Experimental Results Registration phase and Verification phase are the two phases which are available to the user. Registration phase includes two sub phases which are creating new user and recording gesture. New users along with his/her name and user id are created as part of registration phase. We have restricted the number of users to 22 keeping speed in mind. Once the user account is created, the user is allowed to draw gesture using accelerometer and that gesture is the stored in the database in an unreadable format. All users whose accounts have been created will store their gesture in the database. While drawing gesture the user is supposed to give his user id as input along with gesture. Figure 4: Authentication is success and user name is John 4121

5 When a registered user gives incorrect gesture or when a user who is not at all registered draws gesture for verification, then the system will show authentication failure without displaying user name. It is shown in fig 7 and fig 8. Figure 5: Authentication is success and user name is Shan Figure 7: Authentication is failure and no user name is displayed Figure 6: Authentication is success and user name is Akash Figure 8: Authentication is failure and no user name is displayed 4122

6 Conclusion By hand motion gestures portable device generate gestures. High recognition accuracy is provided bt frequency and time domain s of motion recognition s acceleration signals. Computer will receive the data sent by accelerometer that measures acceleration made by hand signals. The overall gesture recognition rate is found to be good compared to existing methods. This results enthusiasts us to additional complicated gestures as user input in various applications. Additional information is provided to the users which indicates starting point and ending point of motion. [10] L. Bao and S. S. Intille, Activity recognition from user-annotated acceleration data, Pervasive, Lecture Notes in Computer Science, no. 3001, pp. 1-17, [11] Y. P. Chen, J. Y. Yang, S. N. Liou, G. Y. Lee, and J. S. Wang, Online classifier construction algorithm for human activity detection using a triaxial accelerometer, Appl. Math. Comput., pp , Nov [12] L. Wang, Feature selection with kernel class separability, IEEE Trans. Pattern Anal. Mach. Intell., vol. 30, no. 9, pp , Sep References [1] S. Zhou, Q. Shan, F. Fei, W. J. Li, C. P. Kwong, and C. K. Wu et al., Gesture recognition for interactive controllers using MEMS motion sensors, in Proc. IEEE Int. Conf. Nano/Micro Engineered and MolecularSystems, Jan. 2014, pp [2] S. Zhang, C. Yuan, and V. Zhang, Handwritten character recognition using orientation quantization based on 3-D accelerometer, presented at the 5th Annu. Int. Conf. Ubiquitous Systems, Jul. 25th, [3] J. Yang, W. Chang, W. C. Bang, E. S. Choi, K. H. Kang, S. J. Cho, and D. Y. Kim, Analysis and compensation of errors in the input device based on inertial sensors, in Proc. IEEE vol. 30, no. 9, pp , Sep [4] Z. Dong, U. C. Wejinya, and W. J. Li, An opticaltracking calibration method for MEMS-based digital writing instrument, IEEE Sens. J., vol. 10, no. 10, pp , Oct [5] S. Zhou, Z. Dong, W. J. Li, and C. P. Kwong, Handwritten character recognition using MEMS motion sensing technology, in Proc. IEEE/ASME Int. Conf. Advanced Intelligent Mechatronics, 2012, pp [6] J. K. Oh, S. J. Cho, and W. C. Bang et al., Inertial sensor based recognition of 3-D character gestures with an ensemble of classifiers, presented at the 9th Int. Workshop on Frontiers in Handwriting Recognition, [7] Y. Luo, C. C. Tsang, G. Zhang, Z. Dong, G. Shi, S. Y. Kwok, W. J. Li, P. H. W. Leong, and M. Y. Wong, An attitude compensation technique for a MEMS motion sensor based digital writing instrument, in Proc. IEEE Int. Conf. Nano/Micro Eng. Mol. Syst., 2010, pp [8] Z. Dong, G. Zhang, Y. Luo, C. C. Tsang, G. Shi, S. Y. Kwok, W. J. Li, P. H. W. Leong, and M. Y. Wong, A calibration method for MEMS inertial sensors based on optical tracking, in Proc. IEEE Int. Conf. Nano/Micro Eng. Mol. Syst., 2009, pp [9] S. J. Preece, J. Y. Goulermas, L. P. J. Kenney, and D. Howard, A comparison of feature extraction methods for the classification of dynamic activities from accelerometer data, IEEE Trans. Biomed. Eng., vol. 56, no. 3, pp , Mar Authors Profile Dr. V. Ramaswamy obtained his Ph. D. degree from Madras University, in He is working as Professor and Head in the Department of Compute Science and Engineering, SASTRA University, Srinivasa Ramanujan Centre, Kumbakonam, Tamil Nadu, India. He has more the 30 years of teaching experience including his four years of service in Malaysia. He is guiding many research scholars and has published many papers in national and international conference and in many international journals and authored one book. He has visited many universities in USA and Malaysia. He is a life member of Indian Society for Technical Education, India. His research interest includes Fuzzy logic, Cryptography, Data Mining, Image Processing, Video Processing and Pattern Recognition. Parashuram Baraki pursuing Ph. D in Computer Science & Engineering in Jain University Bangalore Karnataka, under the guidance of Dr. V. Ramaswamy and currently working as Associate Professor in the Dept of Computer Science & Engineering in Smt. Kamala & Sri. Venkappa M Agadi College of Engineering & Technology, Lakshmeshwar, Karnataka, India. His research interest includes Image Processing, Video Processing and Pattern Recognition. 4123