Research on Action Recognition Method of Traditional National Physical Education Based on Deep Convolution Neural Network

With the continuous development of machine vision and image processing technology, more and more attention has been paid to human action recognition in physical education teaching. In order to improve the performance of action recognition in traditional P. E. teaching, the method of action recognition based on deep convolution neural network is proposed. The length of elbow joint and shoulder joint was calculated by using the distance between the camera and the action image. According to the range characteristics of sports teaching action, monitoring sports teaching action. Deep convolution neural network was introduced to predict the state variables of PE teaching action, and the coordinate data information of all related nodes was obtained. Based on the theory of Deep Convolution Neural Network, this paper transforms and deals with the action posture of traditional national sports teaching in colleges and universities. Through the probabilistic value of the motion image pixel of the traditional PE teaching in colleges and universities, the motion characteristics of PE teaching are extracted. Through detecting the extreme point of PE teaching action in the scale space, locate the extreme point of PE teaching action range. Using the probability of the range of sports teaching action outside the exercise area, we can identify the traditional sports teaching action in colleges and universities. The experimental results show that the method can successfully identify the traditional national sports teaching behavior. This method has good performance in the precision of motion feature extraction, recognition rate and recognition speed. It perfects the problem of low precision in sports action recognition.