Study on the affective property of music

In this paper, we obtain human psychological data through emotional tests on different kinds of music, and use factor analysis approach to find out three major factors that can truly represent the primary affective properties of music. Based on these three factors, the test music can further be classified into two groups using cluster analysis. By analyzing these music test data using short-time processing and comparing the classification results obtained from cluster analysis, both results are consistent, which indicates that a computer model for perceiving the affective characteristics of music can be established. A verification procedure is also conducted to support our conclusion. © 2005 Elsevier Ltd. All rights reserved.