基于声学特征的语言情感识别  

Speech Emotion Recognition Based on Acoustic Features

作　　者：金琴[1,2] 陈师哲 李锡荣[2] 杨刚[2] 许洁萍[2] 

机构地区：[1]中国人民大学数据工程与知识工程教育部重点实验室,北京100872 [2]中国人民大学信息学院,北京100872

出　　处：《计算机科学》2015年第9期24-28,共5页Computer Science

基　　金：北京市自然科学基金(4142029);中国人民大学科学研究基金(中央高校基本科研业务费专项资金)(14XNLQ01)资助

摘　　要：语音情感识别是语音处理领域中一个具有挑战性和广泛应用前景的研究课题。探索了语音情感识别中的关键问题之一:生成情感识别的有效的特征表示。从4个角度生成了语音信号中的情感特征表示:(1)低层次的声学特征,包括能量、基频、声音质量、频谱等相关的特征,以及基于这些低层次特征的统计特征;(2)倒谱声学特征根据情感相关的高斯混合模型进行距离转化而得出的特征;(3)声学特征依据声学词典进行转化而得出的特征;(4)声学特征转化为高斯超向量的特征。通过实验比较了各类特征在情感识别上的独立性能,并且尝试了将不同的特征进行融合,最后比较了不同的声学特征在几个不同语言的情感数据集上的效果(包括IEMOCAP英语情感语料库、CASIA汉语情感语料库和Berlin德语情感语料库)。在IEMOCAP数据集上,系统的正确识别率达到了71.9%,超越了之前在此数据集上报告的最好结果。Emotion recognition from speech is a challenging research area with wide applications. This paper explored one of the key aspects of building an emotion recognition system： generating suitable feature representation. We extracted features from four angles： （1） low-level acoustic features such as intensity, F0,jitter, shimmer, spectral contours etc. and statistical functions over these features, （2） a set of features derived from segmental cepstral-based features scored against emotion-dependent Gaussian mixture models, （3）a set of features derived from a set of low-level acoustic code- words, （4）GMM supervectors constructed by stacking the means or covariance or weights of the adapted mixture com- ponents on each utterance. We applied these features for emotion recognition independently and jointly and compared their performance within this task. We built a support vector machine（SVM） classifier based on these features. We tested the performance of these different features on some public emotion recognition corpus（including IEMOCAP corpus in English, CASIA corpus in Mandarin, and BerlinEMO-DB in Germany）. On the IEMOCAP database, the four-class emo- tion recognition accuracy of our system is 71.9M,which outperforms the previously reported best results on this data- set.

关 键 词：语音情感识别 声学特征 特征融合 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]