Interval Based Speed Invariant Gait Recognition

Abstract

recognition. For each normalized silhouette, the outer contour is constructed by considering only the leftmost and rightmost pixels. Then, distance vector is generated by computing distances between the leftmost and rightmost pixels. The same technique is applied to all silhouettes of an individual. At the end, feature matrix is obtained by concatenating the distance vectors in row wise. Further, multi interval features are generated by considering the each column of feature matrix. In this work, a new similarity matching technique between the gallery and probe interval is proposed. Finally, nearest neighbor classifier is used for the classification procedure.

Keywords: Distance vector; Human gait; Intervals; Outermost contour; Similarity