Random Forest and PCA for Self-Organizing Maps based Automatic Music Genre Discrimination

Digital music distribution industry has seen a tremendous growth in resent years. Tasks such us automatic music genre discrimination address new and exciting research challenges. Automatic music genre recognition involves issues like feature extraction and development of classifiers using the obtained features. As for feature extraction, we base on Self-Organizing Maps to map the high-dimensional audio signals into SOM features. In addition we use Principle Components Analysis (PCA) to reduce feature dimension to improve the classification performance. Regarding the task of genre modeling, we introduce a new method Random Forest. Experiment results show that SOM feature is feasible for music genre classification. Comparisons with traditional classification methods show that the new introduced method can achieve the highest recognition rate. In addition we find that PCA can further improve the discrimination performance of almost all the classifiers we investigate.