Dynamic gesture recognition based on kinect skeleton data

As it is hard to fulfill the hand tracking and gesture recognition tasks in complex environment using a common camera, this paper presents a dynamic gesture recognition method based on skeleton data. Firstly, we get the skeleton joints’ actual position from depth image obtained by Kinect sensor. Then, joints’ trajectories are extracted as the feature vectors. Finally, dynamic time warping (DTW) algorithm is used to achieve fast dynamic gesture trajectory recognition. The method mentioned above is implemented in the human-computer interaction field for the operation of PPT presentation. Experiment results show that the method has good robustness in complex backgrounds and different illumination conditions and the average recognition rate reaches 97.2%.