Music-driven Dance Motion Synthesis

We introduce a novel method for synthesizing dance motions to closely follow the emotions and contents of the input music. The key of our music driven dance motion synthesis method is to model the motion to music mapping relationship exhibited in the sample dance motions and their accompanying background music through a learning based approach. To establish the motion to music mapping relationship, we train a motion to music matching quality rating function according to the sample dance data. To generate an optimal sequence of dance motion segments to match with a piece of input music, we introduce a constraint based dynamic programming procedure. This procedure considers both the motion to music matching quality and the visual smoothness in the resultant dance motion sequence. To evaluate the effectiveness of our method, we quantitatively compare the dance motions synthesized by our method with motion synthesis results by peer methods with respect to the dance motions by professional human dancers in reality. We also conduct several medium-scale user studies to explore how perceptually our dance motion synthesis method can outperform existing methods in synthesizing dance motions to match with the input music. These user studies have produced very positive results.