A Multimodal Complex Model for Measuring the Quality of Music Performance

While computational aesthetics evaluation has achieved success in various domains, the exploration of music performance remains an untouched area. Currently, subjective evaluation stands as the ultimate approach in music aesthetics research, yet it demands substantial human and material resources. Moreover, AI-generated music performances tend to be mechanical, monotonic, and devoid of beauty. To direct the creation of AI music performances and enhance the performance quality of human-being artists, this article introduces an objective beauty measurement method based on Birkhoff's aesthetic measure. The article's primary contributions are twofold: firstly, we present an objective aesthetic evaluation method tailored for music performance; secondly, we introduce ten fundamental music features and four aesthetic music features. Experimental results demonstrate that our method excels in performance assessment.