Multi-Feature Fusion Model for Accurate Dance Movement Recognition Using Spatiotemporal Features and Deep Learning

With the development of multimedia and Internet technology, the demand for precise movement recognition in dance teaching and other fields is increasing day by day. This study presents basic theories and techniques of computer vision and image processing, such as deep learning models and feature extraction methods. According to the characteristics of dance videos, a multi-modal fusion model combining spatiotemporal features is proposed to effectively capture the temporal continuity and spatial distribution characteristics of dance movements. Experiments show that the model outperforms existing methods on multiple public data sets, especially under complex background and illumination changes, and it can still maintain high recognition accuracy, which provides a new idea for the development of dance video motion recognition technology. Experiments show that computational vision can achieve about 95% accuracy in dance movement recognition, and at the same time, it can effectively preserve and recognize images and eliminate invalid movements in key frames of video. The convolutional neural network can achieve a recognition score of 2.4 for the head edge change in the action space score and an efficient recognition speed of up to 1.75Formula: see texts. At the same time, it ensures the classification effect while greatly reducing the number of parameters, improving the training efficiency and improving the recognition accuracy.