摘要
音乐类型分类是音乐检索中非常重要的一个方面。采用支持向量机方法进行音乐类型分类,提取B样条小波特征作为音乐的特征。采用指数径向基函数(ERBF)内核,分类正确率可达86%,比传统的混合高斯模型和K近邻分类器,分类性能分别提高了22%和24%。实验结果表明,采用B样条小波和支持向量机方法是一种有效的音乐类型分类方法。
Musical genre classification is essential in music retrieval. The SVMs method is applied to musical genre classification, and B- spline wavelet feature is extracted as the features of music. Exponential Radial Basis Function （ERBF） kernel function is used to classify the musical genre,86% of classification correctness rate is achieved. In comparison with Gaussian Mixture Model （GMM） classifier and K Nearest Neighbouring （KNN） classifier,the performance of this classification improves 22% and 24% respectively. Experimental results indicate that the method using B-spline wavelet and SVM is effective for musical genre classification.
出处
《计算机应用与软件》
CSCD
2009年第11期221-222,245,共3页
Computer Applications and Software
关键词
音乐类型分类
小波
支持向量机
核函数
Musical genre classification Wavelet Support vector machines Kernel function
