Copyright © 2018 Authors. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. International Journal of Engineering & Technology, 7 (4.7) (2018) 127-130 International Journal of Engineering & Technology Website: www.sciencepubco.com/index.php/IJET Research paper Gait Recognition as Non-Intrusive Biometric Using View Invariant Methods in Multi Temporal Images D. BEULAH DAVID 1 , M.A.DORAIRANGASWAMY 2 1 Research Scholar, Department of Computer Science And Engineering, Sathyabama University, Chennai, India 2 Professor, ASIET, Kalady, Kerala *Corresponding author E-mail: Abstract Gait patterns have been used widely in recent years to authenticate users. Because it doesn’t require user intrusion, it is o ften used as a biometric to make authentication processes easier and hassle free. But there are various issues with this process. To start with, the view- ing angle has to be constant which is quite difficult to achieve with limited number of cameras. Speed too can alter the way a person walks and cause inconsistencies in identification. Furthermore, complications might arise if the subject is carrying something. The weight can affect his walking pattern. Besides, a recent accident could also transform a person’s walking pattern and lead to wrong identification. Other biometrics such as face detection can be combined with this technique to reduce the issues leading to erroneous identification. In this paper, we propose a system to overcome the viewing angle discrepancies. The system takes in walking sequences as input and pro- cesses them to create images. This is converted into 3D images by means of stereovision algorithms. Using which, we can effectively match the real-time image with various image sequences in the database. Side face detection can enhance the accuracy further.. Keywords: Background Subtraction, Transformation, Gait Feature Extraction. 1. Introduction Gait techniques have often been used to identify people based on their walking pattern. But they are not yet employed commercially because of the various issues that arise starting from the person’s clothing to his walking speed. However, there are some instances where the gait technique has been successfully used. In one such case, a burglar was caught using Gait. The footage available from the scene wasn’t clear enough for a facial identification but his walking pattern was adequate. This was further fed into the system and the man was identified and arrested. But since it highly im- portant to identify the right suspect, we cannot allow even the smallest miscalculations. Hence we need a robust software which would reduce the transformation errors to the maximum. Also facial recognition and foot pressure identification is used in addi- tion with gait to slim identification faults. Gait recognition is a biometric method for recognizing people’s walking pattern without intrusion. Patterns can be recognised from a distance from even low quality videos, making them much more efficient than other biometric mechanisms. But there are limita- tions to employing them in real life. The most important being the distortions in views that generally occur in real situations. The most challenging task here is to match the pattern across diverse views. We come across two different approaches proposed which will overcome this issue. An appearance-based approach [14], and a model-based approach [15]-[19]. Appearance-based mechanisms extract gait attributes directly from image sequences captured. On the other hand, model-based mechanisms derive model attributes from images. 3D model-based mechanisms are preferred because of their view-unvarying nature [17], [19], [20]. But it is often chal- lenging to generate 3D modelsof high certainty from images taken using just surveillance cameras.Therefore, we concentrate on ap- pearance-basedmechanisms in this paper.Many methods have been suggested to overcome the viewissue in appearance-based approaches [21]–[23]. Broadly, they fall into three groups: view- invariant, visual hull-based,and view transformation-based meth- ods.View-invariant approaches can be categorised as subspace- based, geometry-based, and metric learningbased methods. Except for the view invariant approach, the remaining approaches use discrete views which are contained in the training set alone and can affect the accuracy if target views aren’t from the trainingsets. View transformation methods extended to arbitrary views solve the discretion issues.3D gait sequences of multiple training views are taken and the features of the target subjects are created by projecting the sequences into 2D spaces related to the target views. The difference is employing multiple non target cases instead of target images. Also, we employ part dependant view selection which separates the gait characteristics along several body parts to fix destination views for each body part. 2. Existing Systems Shuai Zheng et al [1] offered a solution for viewing angle varia- tion issues. The gait energy image was used to create a robust view transformation model. The features were extracted from the energy image using the partial least square method. It eventually performed better by remaining robust to clothing, viewing angle variations and carrying condition changes. Xiaoli Zhou et al [2] utilised the side face of a person and his walking pattern. The side angle of face isn’t usually of high reso-