Music classification using the bag of words model of modulation spectral features

In this paper, the bag of words (BoW) representation of modulation spectral analysis of spectral as well as cepstral features will be constructed for music genre classification. First, the modulation spectrum of each spectral and cepstral feature will be obtained through modulation spectral analysis of a longer texture window. Then, the BoW model of the modulation spectral features will be constructed and used for music classification. Experiments conducted on the music database employed in the ISMIR2004 Audio Description Contest have shown that the proposed BoW representation of modulation spectral features can achieve promising classification accuracy, particularly when it is combined with the BoW representation of frame features.