Music genre classification by analyzing the subband spectrogram

In this study, music genre classification based on the characteristics of spectrogram was studied. In the proposed method, the capability of multi-resolution analysis of the wavelet package decomposition (WPD) as well as the dimension reduction ability of the singular value decomposition (SVD) was integrated to extract the desired features. Experimental results with the well-known ISMIR 2004 and GTZAN database were used to verify the performance of the proposed method.