Abstract
Automatic music analysis is an increasingly active research area. In this area, two major purposes are to understand and to characterize musical forms because musical form is the organizing principle of music. Uncovering recurring structures and their organization in works offers a way to gain insights into the style or techniques of composers and improvisers. In this paper, an automatic music analysis system based on musical forms is proposed to analyze the structure of music data. In our system, we focus on Rondo and Fugue, which are widely used and strictly structured musical forms. Our system consists of two main stages, i.e. feature extraction and musical form analysis. For feature extraction, we focus on monophonic features and represent music data as melody sequences. For musical form analysis, the compositional rules of Rondo and Fugue are adopted. The analysis results of our system show both the structure and the semantic meanings of the music. We make a series of experiments to show that our system has a high precision of detecting musical forms and analyzing the structure of the music.