Kinect based Frontal Gait Recognition using skeleton and depth derived features

Recognizing humans through gait has been an emanant biometric technology in the recent years owing to the fact that it is unobtrusive since it does not require a subject's cooperation. This paper investigates Kinect based gait recognition of human subjects for surveillance applications especially in narrow corridor and airport scenarios where only the frontal views are available. Two features namely skeleton size feature and projectile motion feature extracted from skeleton data and one feature derived by segmenting the depth data using superpixels followed by SURF descriptor extraction are utilized in a hierarchical framework to obtain the closest matching subject for recognition purposes. The proposed method provides considerable increase in the recognition accuracy and recognition rank in comparison to state-of-the-art gait recognition approaches.