Computational Kinematics of Dance: Distinguishing Hip Hop Genres Provisionally Accepted

Ben Baker^1* Tony Liu² Jordan Matelsky^{2, 3} FELIPE PARODI² John W. Krakauer^{4, 5} Brett Mensh⁶ Konrad Kording^{2, 7}

• ¹Colby College, United States
• ²University of Pennsylvania, United States
• ³Johns Hopkins University, United States
• ⁴Johns Hopkins University, Departments of Neurology, Neuroscience, Physical Medicine and Rehabilitation, United States
• ⁵Santa Fe Institute, United States
• ⁶Howard Hughes Medical Institute, Janelia Research Campus, United States
• ⁷Canadian Institute for Advanced Research (CIFAR), Canada

Dance plays a vital role in human societies across time and culture, with different communities having invented different systems for artistic expression through movement (genres). Experts can verbalize such differences to an extent, but conveying qualities of movement and kinaesthetic feeling in language has major limitations. Existing dance notation schemes have been instrumental in documenting dance in certain contexts, yet we can do more to capture the important details of movement across a wide spectrum of genres. The field would benefit from a general, quantitative and human-understandable method of characterizing meaningful differences between aspects of any dance style; a computational kinematics of dance. Here we introduce and apply a novel system for encoding bodily movement as 17 macroscopic, interpretable features, such as expandedness of the body or the frequency of sharp movements. We use this encoding to analyze Hip Hop Dance genres, in part by building a low-cost machine-learning classifier that distinguishes genre with high accuracy. Our study relies on an open dataset (AIST++) of pose-sequences from dancers instructed to perform one of ten Hip Hop genres, such as Breakdance, Popping, or Krump. For comparison we evaluate moderately experienced human observers at discerning these sequence's genres from movements alone (38% where chance = 10%), and model baselines line the Ridge classifier (48%). Our model exhibits strong performance (76%), suggesting that the selected features represent important dimensions of movement for expressing of the attitudes, stories, and aesthetic values manifested in these dance forms. Among the features we used, a dancer's expandedness was the most discriminative of genre, but any one or a few features yielded poor performance sufficient to distinguish genre at human-comparable performance. Our study also offers a window into significant relations of similarity and difference between the genres studied, and between the performance of our model as compared to experienced humans. Because of the rich, complex, and culturally shaped nature of the differentiation of Hip Hop Dance genres, the interpretability of our features, and the lightweight techniques used, our approach has great potential for generalization to other movement domains and movement-related applications.