Beat Tracking from Conducting Gestural Data: A Multi-Subject Study

The musical conductor metaphor has been broadly used in the design of musical interfaces where users control the expressive aspects of the performance imitating the movements of conductors. Most of the times, there are predefined rules for the interaction to which users have to adapt. Other works have focused on studying the relation between conductors' gestures and the resulting performance of the orchestra. Here, we study how different subjects move when asked to conduct on top of classical music excerpts, with a focus on the influence of the beat of the performance. Twenty-five subjects were asked to conduct on top of three classical music fragments and recorded with a commercial depth-sense camera. We evaluated predicted beats using ground truth annotations from score-performance alignment by an expert musicologist and a modified F-measure that is able to account for different tendencies on beat anticipation across subjects. The results show that these tendencies can be used for possible improvements in the design of conducting musical interfaces in terms of user adaptation.