Boosting Classifiers for Music Genre Classification

Abstract

Music genre classification is an essential tool for music information retrieval systems and it has been finding critical applications in various media platforms. Two important problems of the automatic music genre classification are feature extraction and classifier design. This paper investigates discriminative boosting of classifiers to improve the automatic music genre classification performance. Two classifier structures, boosting of the Gaussian mixture model based classifiers and classifiers that are using the inter-genre similarity information, are proposed. The first classifier structure presents a novel extension to the maximum-likelihood based training of the Gaussian mixtures to integrate GMM classifier into boosting architecture. In the second classifier structure, the boosting idea is modified to better model the inter-genre similarity information over the mis-classified feature population. Once the inter-genre similarities are modeled, elimination of the inter-genre similarities reduces the inter-genre confusion and improves the identification rates. A hierarchical auto-clustering classifier scheme is integrated into the inter-genre similarity modeling. Experimental results with promising classification improvements are provided.

This work has been supported by the European FP6 Network of Excellence SIMILAR (http://www.similar.cc). The authors would like to thank George Tzanetakis for sharing his music genre database with us.