Design and implementation of an intelligent evaluation and correction method for dance postures

To enhance motion recognition accuracy and posture correction efficiency in dance training, an intelligent system integrating posture recognition, error evaluation, and visual feedback was developed. This study investigates a high-resolution network-based skeleton keypoint extraction method, combined with dynamic time warping and a multidimensional error fusion model, to achieve efficient alignment and precise evaluation between standard movements and real-time postures. Experimental validation using ballet, folk dance, and modern dance motion samples demonstrates the system's superior performance in keypoint recognition accuracy, angular error correction, and feedback response speed, showcasing strong real-time capability and stability.