In order to quickly and accurately identify dynamic gestures of train driver

a dynamic time warping（ DTW） algorithm based on machine vision is proposed. Kinect vision sensor is employed to extract depth information of gesture

and gesture information is segmented from the background through selecting appropriate depth-distance threshold

with the combination of the information of human skeleton node. The segmented gesture image is identified by support vector machine（ SVM） algorithm

and the normalization for different gestures is also evaluated. The motion sequence of arm skeleton node is derived depth image and the driver’s bone data

and the optimal matching of dynamic gesture is implemented with DTW algorithm. The recognition result and evaluation for the dynamic gesture of train driver is obtained. The method is tested with the data from simulation scenes

and the results show that the proposed method can give better result than traditional methods in the reliability and stability. In addition

the method can satisfy real-time recognition and give an accurate evaluation score.