Optimization of locally linear embedded for frontal gait recognition using kinect

Abstract

This study investigates the potential of gait features as human gait recognition. Firstly, skeleton joints of twenty subjects obtained from Kinect are extracted as features and further selected using the optimized locally linear embedded approach. Next, multi-layer perceptron and support vector machine are employed as classifiers. Result showed that the combination of the optimized locally linear embedded with K=100 and d=94 and support vector machine regularization parameter C=0.001 and linear kernel attained highest accuracy rate in frontal view specifically 96.50% using 94 gait features.