Dance Teaching Instruction Method Based on Image Movement Segmentation and Deep Learning

Many dance teachings still rely on traditional face-to-face demonstration and imitation, lacking innovation and diversity. This approach may not fully stimulate students’ interest and motivation, leading to poor learning results. Therefore, a dance teaching and guidance method based on image movement segmentation and deep learning is proposed. By combining the image preprocessing technology of Radon transform and wavelet multi-scale decomposition, the trajectory of dance jump movement is extracted in real time. The dance image is segmented. The frequency domain features of dance movement contour are obtained through Fourier transform. The time dependence and context information of dance movement are captured through the long short-term memory network. The scale features at different levels are extracted and attention graphs are generated. Based on the comparison of image features, the correct recognition and correction of wrong actions are realized, and a comprehensive teaching guide scheme is given. The experimental results show that the average recognition rate of the design method is [Formula: see text]%, the goodness of fit value is high, the extraction time is [Formula: see text][Formula: see text]s, and the image peak signal-to-noise ratio is higher than [Formula: see text][Formula: see text]dB.