线性判别分析和支持向量机的音乐分类方法

提出了一种新的音乐分类方法,该方法使用线性判别分析(LDA)和支持向量机(SVMs)对音乐数据进行分类。在实现音乐分类中,先使用傅里叶变换等方法从每一段音乐中提取音频特征,包括Mel倒谱系数及基音频率等,并将它们按比例组成一个高维向量;再使用LDA对这些高维向量进行降维,使得各类音乐的类间离散度与类内离散度的比值最大;最后使用SVM等4种分类器对降维后的特征进行分类。实验证明LDA及SVM使得音乐分类的精确度有了较大的提高。

Music Classification Based on Linear Discriminative Analysis and Support Vector Machine

In this paper, a new music classification method is presented. The LDA(Linear Discriminative Analysis) and SVMs(Support Vector Machine) are used in this method to classify the music data accurately. DFT is applied to extract several basic audio features, including the Mel-frequency cepstral coefficients and pitch frequency which are concatenated to form a high-dimensional vector. The dimensionality reduction is performed on the vector. Finally, 4 classifiers(including SVMs) are used to classify the music. The experimental results show that the LDA and SVM improve the accuracy of the classification greatly.