论文部分内容阅读
为了提高基于回归的音乐情感分类准确率,文中运用了k平面分段回归的方法,在音乐特征与音乐情感组成的高维空间内,通过多次迭代寻找超平面的方法直接求解非线性回归问题,进而预测二维情感变量值Valence与Arousal,并通过该二维情感变量值进行音乐情感分类。为了验证分类系统的性能,实验中按MIREX分类标准建立有5类音乐情感的音乐库,对其300首音乐样本进行分类,与传统的多元线性回归和支持向量回归相比分类准确率有了一定提高。表明k平面分段回归的方法可以有效运用于音乐情感分类。“,”In this paper,a piecewise regression approach of k-plane is employed in order to improve the classification accuracy of music emotion based on regression. It solves the nonlinear regression problem directly through several iterations in high dimensional space con-sisted by music feature and music emotion,predicting the valence and arousal values in the emotion model and classifying the music emo-tion. To verify the performance of classifier,test the classifier on 300 music sample from a music dataset which is commonly employed in MIREX,and the testing results on classification accuracy of the proposed approach are compared with the results from multiple linear re-gression and support vector regression. The experimental results show that k-plane piecewise regression approach can achieve the higher accuracy than the other two. That is to say the method of k-plane piecewise regression can be effectively applied to music emotion classi-fication.