Physical Education’s Role in Public Health based on the combination of speech recognition and behavior capture

Abstract

Based on speech recognition and behavior capture, human motion correction is designed. In the process of movement, the real-time movement change of human posture is recognized, and the result of movement correction analysis is given after the movement. In the process of motion capture and correction, the position data of human bone joints should be taken into account, and the adjacent joints should be used to form the direction of the limb to carry out real-time motion correction. Specifically, according to the standard action corresponding to the joint point coordinates of the action to be detected, when the joint point bone data and the estimated value of the action to be detected are within the allowable range, it is judged to be in line with the standard action; on the contrary, according to the symbol of the error between the joint point and the estimated value, the deviation direction of the action to be detected is judged and corrected. Because the difference of action sequence can be regarded as the difference of feature in action sequence, the appropriate feature is selected to describe the process of motion difference according to the action sequence composed of bone joint position data. For action comparison analysis, considering the existence of action sequences with different lengths, firstly, the dynamic time warping method is used to align the action sequence; then the feature vector of the aligned action sequence is extracted, the cosine similarity is used to judge the action similarity, and the angle feature is used to describe the motion trajectory to obtain the result of action comparison analysis..