According to the action characteristics of one’s hitting ping-pong ball

propose a recognition algorithm which is based on 3D acceleration signal to distinguish different types of strokes

such as attacking

rubbing and blocking. Data is acquired in training through acceleration sensor tied on wrist; beginning point and end point of acceleration signal are judged by fixed threshold method; and other key characteristics such as signal energy and gesture length are extracted. The K-NN and the decision tree algorithm are used in the process. Result suggests that stroke signals can be classified and distinguished in table tennis training fast and effectively.