Time-Frequency Analysis for Music Genre Classification by using Wavelet Package Decompositions

Abstract

In this paper, time-frequency analysis method was studied in the genre classification of music songs. Due to the benefits of multi-resolution analysis, the proposed methods first adopted the wavelet package decomposition (WPD) to obtain candidate features from the spectrograms of music songs. Then the singular value decomposition (SVD) was used to extract the desired features because of its dimension reduction ability. In the experiments, well-known database named ISMIR2004 was applied. To examine the performance of the proposed methods, the experimental results of the proposed methods were compared with the results of the MFCC based methods.