结合韵律和音质参数的改进二次判别式在语音情感识别中的应用  

Speech Emotion Recognition Using Prosody and Voice Quality Features by Modified Quadratic Discrimination Function

作　　者：赵艳[1] 赵力[1] 邹采荣[1] 

机构地区：[1]东南大学信息科学与工程学院,江苏南京210096

出　　处：《信号处理》2009年第6期882-887,共6页Journal of Signal Processing

摘　　要：提取可以表征唤醒维的韵律参数和表征效价维的音质参数综合用于语音情感识别是近年来此领域的一个研究方向。为了提高情感识别率,本文提取了18个韵律参数和59个音质参数用于识别,为避免特征矢量维度过高而造成的计算量过大和信息冗余,采用主分量分析神经网络(PCANN)进行降维,并用二次判别式进行参数有效性验证。针对二次判别式输入参数正态化这一假设,提出一种改进二次判别式用于四种情感的识别,实验结果表明,改进方法可以有效提高识别率。One focus in speech emotion recognition is the utilization of prosody and voice quality features. In this paper, 18 prosody features in arousal dimension and 59 quality features in valence dimension are extracted, and principal component analysis neural network （PCANN） is used to reduce dimension of the feature vectors. Quadratic discrimination function （QDF） is introduced to verify the features validity. Then, a modified quadratic discrimination function （MQDF） is proposed to normalize the input features; the resuits show that the modified method in this paper could improve the recognition ratio effectively.

关 键 词：语音情感识别 唤醒维 效价维 音质参数 PCANN 改进二次判别式 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]