Real-time Autonomous Dancing Robot System based on Convolutional Neural Network
Min-Geun Cho, Hong Seong Park · Journal of Institute of Control Robotics and Systems · 2017
This paper proposes a real-time autonomous dancing robot system that listens to arbitrary music inputted in real time and then can dance according to the genre of the music. The proposed system autonomously performs the whole dancing by selecting the dancing motion that matches the music genre and predicting the beat-related information for synchronization between the robot motion and the music. The proposed system updates the data of the input audio signal every 5 second using the first-in first-out method. In addition, the updated data are used as the input for both the short-term music genre classifier based on CNN (Convolutional Neural Network) and the beat detector to detect the beat interval time and the beat start time. A rule-based decision algorithm is suggested to improve the decision success rates of both the genre classifier and the bit detector and to detect the change of music genre simultaneously. Also the proposed system is verified by some examples.