Research and application of running action sequence recognition algorithms based on kinect

Abstract

In order to obtain better analysis of human characteristics thus human-computerinteraction, recognition algorithms based on skeletal binding of running, waving, jumping and crouching were presented in this paper. It was based on characteristics analysis of human motion and threshold setting. The human skeleton data obtained from Kinect sensors can be used to bind the human skeleton with its corresponding computer model. The algorithms were also applied in a new developed treadmill system based on Kinect. Interactive running (fitness entertainment sports) was implemented with this system based on recognition of human actions. The experimental data shows the effectiveness of the running sequences recognition algorithm based on skeleton binding, with a recognition accuracy up to 92% which can fully meet the requirements of somatosensory game functions. The research results can be potentially used for other human-computer interaction application areas.