Enhancement of Innovativeness of Dance Works by Artificial Intelligence-Based Dance Creation System

Abstract

The gradual development and maturation of artificial intelligence technology has driven its deep integration and application in the field of dance choreography. This paper proposes a music-dance generation model based on generative adversarial networks (GANs), which consists of a generator and a discriminator. The generator employs the CL-K2M model, while the discriminator determines whether the generated dance sequence is authentic under the condition of known music. Additionally, the model calculates the pose differences between two frames based on the differences in degrees of freedom, and combines dance movement characteristics with motion interpolation techniques to achieve smooth transitions between two sequences. The dance movements generated by the designed model achieved the highest average user ratings in three metrics: rhythm (7.57), coordination (7.35), and style retention (8.45), demonstrating significantly superior performance compared to similar model algorithms.