Music genre/mood classification using a feature-based modulation spectrum

The feature-based modulation flatness measure (FMSFM) and feature-based modulation crest measure (FMSCM) are proposed as novel feature vectors for music genre and mood classification. These features are extracted using a feature-based modulation spectrum to represent time-varying characteristics of the music signal. Instead of the spectrogram of the signal, timbral features such as mel-frequency cepstral coefficient (MFCC), decorrelated filter bank (DFB), and octave-based spectral contrast (OSC) are used for modulation. Combining FMSFM and FMSCM with the timbral features, we obtain significantly better accuracy in genre and mood classification than the conventional features.