Real-time Beat Tracking System with Mel-Frequency Cepstral Coefficient Method and ANN for ERISA Dance Robot Motion

The Indonesian Robot Contest (KRI) is a robotics-related design and engineering competition. One of the competition categories is the Indonesian Dance Robot Contest (KRSTI), which requires contestants to create robots that can perform to the beat of the music. ERISA is a humanoid robot created by PENS students for the competition. One of the evaluation criteria for the competition is the congruence between dance movements and musical accompaniment. This paper focuses on designing a system for humanoid robots to be able to dance according to certain parts of music. If the robot has certain troubles during performances, it is not necessary to repeat the motions from the beginning instead of following the beat at that time. The proposed system combines the Mel Frequency Cepstral Coefficient for feature extraction and the Neural Backpropagation Network for voice recognition. The effectiveness was tested through experiments for music recognition, which indicate that the robot’s average response time is approximately 13.011 seconds, with a success rate of 100 percent in determining movement.