A real-time music performance emotion recognition system integrating computer vision and audio signal processing

With the rapid development of intelligent technology, emotion recognition has become an important research direction in human-computer interaction, intelligent entertainment and other fields. This paper proposes a real-time music performance emotion recognition system that integrates computer vision and audio signal processing. The system recognizes the emotional state in music performance in real time by combining visual information (e.g., performer's facial expression and body movement) and audio signals (e.g., pitch, rhythm, timbre, etc.). The system uses computer vision technology to analyze the performer's facial expression and body movements to extract emotional features; the emotional information of music is extracted through audio signal processing. The deep learning model is used to fuse and analyze the multimodal data to identify the emotional categories in the performance. The experimental results show that the proposed system is able to realize high accuracy emotion recognition in complex performance environments with strong real-time and robustness. This paper discusses the prospects for the application of the system in the fields of music entertainment, education and psychology, and proposes future research directions.