Recently considerable research has been conducted to retrieve pertinent parameters and adequate models for automatic music genre classification using different databases. Many of previous works are derived from speech and speaker recognition techniques. In this paper, four measures are investigated for mapping the features space to decision space. The first two measures are derived from second-order statistical models and last measures are based upon information theory concepts. A Gaussian Mixture Model (GMM) is used as a baseline and reference system. For all experiments, the file sections used for testing have never been used during training. With matched conditions all examined measures yield the best and similar scores (almost 100%). With mismatched conditions, the proposed measures yield better scores than the GMM baseline system, especially for the short testing case. It is also observed that the average discrimination information measure is most appropriate for music category classifications and on the other hand the divergence measure is more suitable for music subcategory classifications.
展开▼
