A Framework for Multi-Label Music Mood Classification

Abstract

Music is the effective communication medium among people. Studying music mood can help in music understanding, music retrieval, and some other music-related applications. This paper presents a hierarchical framework with a new mood taxonomy model to automate the task of mood classification from acoustic music data based on western music psychology theory. This system proposes hierarchical framework with new mood taxonomy model. The proposed mood taxonomy model is combined by the Thayer’s 2 Dimension model and Schubert’s updated Hevner adjective checklist. The 60 famous English songs are used as the standard database in this system which is created by literature. The verse and chorus part from the whole song is extracted manually for processing in this proposed system. The extracted music clip is segmented by image region growing method to separate homogenous part on the entire music clip. Then, the feature sets from the separated music trimmed are extracted to inject the Fuzzy Support Vector Machine (SFVM). To solve the multi-label classification problem, one-against-one (O-A-O) multi class classification method are used. The hierarchical framework with new mood taxonomy model has the advantage of reducing the number of classifier used for O-A-O approach.