Design and Testing of a Generic Algorithm for Assessment of Human Walking

This paper addressed a novel scheme of training Support Vector Machines (SVMs) for automatic discrimination of the change of human walking patterns. In order to classify gait patterns with higher accuracy, the independent component analysis (ICA) algorithm was proposed as a new pre-processing technique to extract gait features for initiating the training set of SVM. The gait data of 30 young and 30 elderly participants were acquired during normal walking, and were analyzed using the proposed method. The test results indicated that the ICA algorithm could obtain more gait features containing high-order statistical information about the change of young-old walking pattern, and ICA-based SVM has achieved an evidently improved the generalization performance, compared with PCA-based SVM. The proposed model could be functioned as an efficient tool for the assessment of the human walking in the future clinical application.