Dynamic Gait Recognition of Chinese Dance Based on Contour Features

Abstract

Single feature recognition has certain limitations, which can't fully reflect the differences between dance gait information, resulting in low recognition rate of frame difference features. In this paper, a dynamic gait recognition method for Chinese classical dance is proposed based on contour features. The gait image of Chinese dance is preprocessed, the target region is segmented, and the gait cycle is extracted. The gait energy map is composed of the sequences in a cycle, which contains the information of the upper and lower limb motion frequencies of the dancer. The key distance, width, wavelet features and gait energy map are fused to form a group of feature vectors. The problem of low efficiency of single feature recognition is solved by feature fusion. Based on CNN model, a dynamic gait recognition model of Chinese dance is established based on contour features. Under the function of convolution kernel, the values of each channel are concatenated into a one-dimensional vector to express the contour features and complete the video sequence classification. The test results show that the design method retains the correlation between the gait data of Chinese classical dance, accurately locates the key points, and reflects the movement and time information of gait, so the recognition rate is improved.