Regularized eigenspace-based gait recogntion system for human identification

Gait has received much interest from the biometric society in the vision field due to its utility in walker identification. In this paper, we present a regularized eigenspace-based gait recognition system for human identification. First, motion contour image (MCI) is extracted from walking sequences. In training phase, eigenfeature regularization and extraction (ERE) is applied to the gallery motion contour image to obtain regularized transformation matrix and gallery features. In test phase, regularized transformation matrix is applied to project motion contour image into the eigenspace to obtain probe features, and determine the identity based on the result of nearest neighbor classifier. Experiments are performed with the CASIA gait database A to evaluate the performance of the proposed gait recognition system. Experimental results justify the superiority of our proposed system in terms of correct classification rate.