Music Classification using Multiclass Support Vector Machine and Multilevel Wasserstein Means
Wei Jin, Cong Jin, Zhiyuan Cheng, Xin Lv, Leiyu Song · 2019
Music classification is a challenging task in music information retrieval. In this article, we compare the performance of the two types of models. The first category is classified by Support Vector Machine (SVM). We use the feature extraction from audio as the basis of classification. Firstly, a total of 500 pieces of music by five famous classical music composers were selected, 400 of which were regarded as the training set of music genre classification, and the remaining pieces were regarded as the testing set. The second method is Multilevel Wasserstein Means(MWMs). From the experimental results, Multilevel Wasserstein Means, as an unsupervised learning, has been able to approach SVM in classification results,and has achieved 85% accuracy.