基于多级SVM分类的语音情感识别算法

为了提高语音情感识别系统的识别准确率,本文在传统支持向量机(SVM)方法的基础之上,提出了一种基于PCA的多级SVM情感分类算法。首先将容易区分的情感分开,针对混淆度大且不能再利用多级分类策略直接进行区分的情感,采用主成分分析法(PCA)进行特征降维,然后逐级地判断出输入语音所属的情感类型。与传统基于SVM分类算法的语音情感识别相比,本文提出的方法可将7种情感的平均识别率提高5.05%,并且特征维度可降低58.3%,从而证明了本文所提出的方法的正确性与有效性。

A speech emotion recognition algorithm based on multi-layer SVM classification

In order to improve the accuracy of speech emotion recognition systems, based on the traditional support vector machine (SVM) method, this paper proposes a multi-layer SVM emotion classification algorithm combined with PCA. The first layer separates the emotions which can be easily distinguished. For the emotions which can"t be distinguished by the first layer, feature dimension reduction is performed with the principal component analysis (PCA) method, and then emotions are separated layer by layer. Compared to the traditional SVM classifier, the proposed PCA-combined multi-layer SVM classifier improves the average recognition accuracy of 7 emotions by 5.05%, and the feature dimension are reduced by 58.3%, thus proves the correctness and effectiveness of the proposed method.