Research On Recognition Of Stroke Movement In Table Tennis Based On Acceleration Sensor

Action recognition as a kind of important technology of human computer interaction, it is a hot field of pattern recognition. The acceleration sensor is a kind of wearable sensor. With low energy consumption, small size and low cost, it is widely used.At present, the process of motion recognition based on acceleration sensing is divided into 3 steps: feature extraction, feature selection and recognition algorithm. In this paper, a recognition algorithm based on 3D acceleration signals is proposed to recognize different types of shots, which is based on the motion characteristics of a person who hit the table tennis. For example: to attack the ball, rub the ball and blocking. The data is collected by the acceleration sensor which is bound on the wrist in the course of the table tennis training. Using the fixed threshold method to judge the acceleration signal of the beginning and end point, and extracts key features of the signal energy, gesture length of the signal and the like. The k nearest neighbor decision tree algorithm and support vector machine algorithm are used to identify the table tennis balls. Dynamic time warping algorithm Tennis Hitting online identification work, and make a comparative analysis of experimental results of different algorithms.By comparing and analyzing the experimental results of different recognition algorithms in the recognition system of table tennis ball is based on acceleration sensor, the results show that the ball movement in the course of table tennis training.In this paper, the recognition algorithm of off-line and online two cases can be quickly and efficiently to classify and recognition of motion signals. It reflects the scientific and rationality of the identification model.