Wavelet analysis for audio signals with music classification applications

Audio data like speech and music can be analyzed and processed with Fourier methods, having as one constraint the constant product of time and frequency resolutions. This problem can be avoided applying the wavelet transform, ensuring good resolutions on both time and frequency supports. We propose in this paper to determine features of music in a combined framework using multi-resolution (wavelet) analysis and spectral analysis in order to realize the classification of musical pieces in genre classes. The proposed approach also uses a number of features commonly employed for speech recognition, such as Mel-cepstral coefficients, zero crossing rate or the signal energy. Moreover, the rhythm audio content is considered, the corresponding feature parameters being extracted from beat-histograms.