Research on the Construction of Music Performance Robot Based on Beat Recognition

Abstract

This paper addresses the issue of music performance robots frequently misjudging complex rhythmic patterns, leading to abnormal performance movements. Focusing on rhythm recognition and gesture recognition, we propose an adaptive rhythm identification model based on bidirectional recurrent neural networks and gesture recognition technology for a performing robot system. Experimental results demonstrate the model's superior performance in robot dance performances with complex musical rhythms. Specifically, the bidirectional recurrent neural network adaptive model enhances the accuracy of rhythm recognition to over 90%, while gesture recognition technology matches the rhythm identification results to dance movements through complete information transformation, which are then precisely presented by the robot. Ultimately, experiments confirm that the recognition system meets the design requirements for music performance robots.