A database for aesthetic classification of Chinese traditional music

Abstract

Artificial intelligence has been a research highlight in recent years. Therefore, quantitative analysis about human perception has been of great concern, for instance, affective computing based on music or picture materials. As music is one of the main art forms of auditory aesthetics, its quantitative studies related to aesthetic perception show great potential. This study proposed the evaluation method of aesthetic categories for Chinese traditional music. According to this method, a dataset with aesthetic-emotional multi-annotation was established. The dataset was composed of 500 clips of Chinese traditional music, and it consisted of five aesthetic categories. The distribution characteristics of different aesthetic categories in the emotional dimension space were analysed. Furthermore, by extracting corresponding acoustical features, we tested the accuracy of different classifiers for aesthetic classification. The results showed that the highest accuracy was 65.37% by logistic regression. This work provided a data foundation for the quantitative research and aesthetics computation of Chinese traditional music. The database also can be used for the research of cross-cultural music perception.