Classification of Walking Gait Features using Markerless-based Approach in ASD Children

This paper discussed the gait classification of autism children versus normal group. The study involved children from 30 typically development and 21 autism children with aged range between 6 to 13 years old. In this study, gait data from both groups were captured using markerless approach namely Kinect sensor. Three types of gait features are extracted namely direct joint feature, reference joint feature and center of mass feature. Additionally, all the features are classified using three different types of classifiers. Further, the effectiveness of the features for classification of walking gait pattern for ASD children is evaluated. Based on the results obtained, artificial neural network (ANN) outperformed the other two classifiers and results showed that the direct joint feature contributed to perfect classification followed by reference joint feature and center of mass feature