Anomalous gait detection based on Support Vector Machine

Support Vector Machine is amongst the popular machine classifier due to its rigorous theory background and remarkable generalization performance. Hence, in this paper, the performance of SVM is evaluated to classify gait abnormalities due to Parkinson disease based on three kernels namely radial basis function, polynomial as well as linear. In addition, two types of normalization is applied to these gait data namely intra group norm and inter group norm. Initial findings showed that basic spatiotemporal parameters found to be the most significant features. Results also proven that intra group norm and RBF kernel are capable to to be used in detecting anomaly gait pattern between normal and PD patients based on the accuracy rate attained.