KPCA-based SVM for automated gait patterns recognition

In order to improve the accuracy of recognition of gait patterns, this paper addresses a novel technique using the KPCA-Based SVM algorithm for automated recognition of gait patterns. The proposed technique primarily extracts the nonlinear information of the gait patterns by using KPCA algorithm, which definitely reduces the noise and redundant information. The selected nonlinear information represents the efficient ones which make the training samples of SVM much more representative, thus largely improving the generalization performance of SVM classifiers. With the proposed method, the gait patterns of normal subjects from the young and elderly groups were analyzed. The result shows that the accuracy of gait recognition of our method is 89.6%, which increases by 6.3% and 8.3% respectively, compared with PCA-based SVM method and SVM method.