Abstract: In order to meet the rehabilitation training needs of patients with motor disfunction． A motion model to
extract human motion features is proposed based on the node Angle sequence based and the extraction of human
skeleton topology using Kinect equipment，and a rehabilitation training system based on Kinect motion sensor
camera is designed． The system collect human real-time skeleton data with Kinect，calculate the Angle between
feature nodes，and form the Angle sequence． The DTW ( Dynamic Time Warping) algorithm is used to compare
the similarity between the measured Angle sequence and the standard action sequence of the action library，judge
whether the action is standard and output the evaluation results． It is easy to use and costs less，which improves
the fun of the training process． The system realizes the rehabilitation training guidance for the patients with sports
disorder through human-machine interaction，and has a positive influence on the rehabilitation training process．

Key words: Kinect, rehabilitation training, dynamic time warping ( DTW ) algorithm, humancomputer
interaction