Implementation of a real-time dance ability for mini maggie

Thumbnail Image
Publication date
Defense date
Journal Title
Journal ISSN
Volume Title
Google Scholar
Research Projects
Organizational Units
Journal Issue
The increasing rise of robotics and the growing interest in some fields like the human-robot interaction has triggered the birth a new generation of social robots that develop and expand their abilities. Much late research has focused on the dance ability, what has caused it to experience a very fast evolution. Nonetheless, real-time dance ability still remains immature in many areas such as online beat tracking and dynamic creation of choreographies. The purpose of this thesis is to teach the robot Mini Maggie how to dance real-time synchronously with the rhythm of music from the microphone. The number of joints of our robot Mini Maggie is low and, therefore, our main objective is not to execute very complex dances since our range of action is small. However, Mini Maggie should react with a low enough delay since we want a real-time system. It should resynchronise as well if the song changes or there is a sudden tempo change in the same song. To achieve that, Mini Maggie has two subsystems: a beat tracking subsystem, which tell us the time instants of detected beats and a dance subsystem, which makes Mini dance at those time instants. In the beat tracking system, first, the input microphone signal is processed in order to extract the onset strength at each time instant, which is directly related to the beat probability at that time instant. Then, the onset strength signal will be delivered to two blocks. The music period estimator block will extract the periodicities of the onset strength signal by computing the 4-cycled autocorrelation, a type of autocorrelation in which we do not only compute the similarity of a signal by a displacement of one single period but also of its first 4 multiples. Finally, the beat tracker takes the onset strength signal and the estimated periods real-time and decides at which time instants there should be a beat. The dance subsystem will then execute different dance steps according to several prestored choreographies thanks to Mini Maggie’s dynamixel module, which is in charge of more low-level management of each joint. With this system we have taught Mini Maggie to dance for a general set of music genres with enough reliability. Reliability of this system generally remains stable among different music styles but if there is a clear lack of minimal stability in rhythm, as it happens in very expressive and subjectively interpreted classical music, our system is not able to track its beats. Mini Maggie’s dancing was adjusted so that it was appealing even though there was a very limited range of possible movements, due to the lack of degrees of freedom.
Robotics, Human-robot interaction, Music, Dancing robots, Signal processing
Bibliographic citation