Sensing Dance Engagement for Collaborative Music Control

We introduce a concept that allows attendees of a party to collaboratively influence the music selection process. As explicit feedback is likely to disturb the atmosphere, we introduce an unobtrusive, implicit feedback mechanism. In particular, we propose to sense the partygoers' dance engagement by means of their mobile phones. Since people tend to dance more when they enjoy the music, this metric provides a measure of the crowd's satisfaction. We verify experimentally that music can be selected such that the generated song sequence matches the taste of the crowd, and we demonstrate the feasibility of inferring an individual's dance engagement using the mobile phone's accelerometer. Our experiments show that the proposed method is robust to the body location at which the phone is worn and reaches an overall accuracy of 86.8% for distinguishing dancers from non-dancers.