Gait recognition based on active energy image and parameter-adaptive Kernel PCA

In this paper, we used gait silhouettes that provided by CASIA, and all we study are based on this database. Firstly, we normalized and centralized gait silhouettes and get the gait sequence, secondly, we extract the active regions by calculating the difference of two adjacent silhouettes images, and construct an AEI by accumulating these active regions, finally, using Kernel Principal Component Analysis (KPCA) method to analysis the AEI, and parameter optimization method used to determine the nuclear function of KPCA, and using SVM to classified and recognized gait. Experimental results show that such methods to be identified effective.