Gait Based Person Identification Using Kohonen Self-Organizing Mapping Technique

Abstract

Though gait recognition system has some special security features, it has some limitations like viewing angle variation, slow or fast walk and carrying a bag etc. This paper has an advantage of removing these limitations by using gait energy image with Kohonen self-organizing mapping technique. This model proposes the method that is developed by Partial least square (PLS) used for features selection and singular value decomposition (SVD) used for dimensionality reduction. In existing robust view transformation model (RVTM), the current state can achieve over 80% recognition rate under the conditions where the training and testing data are in similar fashions. But the system performance has been decreased with the change of clothing, shoe, bag, surface and illumination, pose and viewing angles. In RVTM, the average recognition rate under different viewing angles proposed by Shuai Zheng et al. is 51% using the CASIA gait dataset B. In this proposed thesis work, Kohonen self-organizing mapping neural network based model has been used to enhance the performance under various adverse environmental conditions. Experimental results and performance analysis show the results of the proposed system with different viewing angles where the average identification rate has been increased up to 55%.

Keywords: Gait recognition, gait energy image, Kohonen self-organizing mapping technique

Cite this Article

Abu Sayeed, Abhijeet Saha, Md. Rabiul Islam. Gait Based Person Identification Using Kohonen Self-Organizing Mapping Technique. Current Trends in Information Technology. 2015; 5(2):    1–6p.