Music style classification using support vector machine
Jing Lu, Wanggen Wan, Xiaoqing Yu, Changlian Li · 2009
Music style classification is important in numerous research fields. In this paper, a novel method is presented to classify six style music samples. The method makes use of the multi-class support vector machines (SVMs) model based on Mel-frequency cepstrum coefficients to classify different music style files. The "voting strategy" is chosen and "AND" gate is used to combine all of the k(k -1)/ 2 binary classifiers to test samples. After training the data, a model file is created. Then, new input data can be predicted according to the pre-trained model and get the result. The experimental results show that the multi-class support vector machine learning method has fine performance in music classification. Furthermore, the length of the training time and the parameters search using cross-validation are also discussed in this paper.