基于NAQ的语音情感识别研究  

Emotion recognition based on normalized amplitude quotient

作　　者：白洁[1,2] 蒋冬梅[1] 谢磊[1] 付中华[1] 任翠红[1] 

机构地区：[1]西北工业大学计算机学院,西安710072 [2]海军兵种指挥学院作战指挥系,广州510430

出　　处：《计算机应用研究》2008年第11期3243-3245,3258,共4页Application Research of Computers

基　　金：国家自然科学基金资助项目(60703104);陕西省自然科学基础研究计划资助项目(2007F15)

摘　　要：研究了用迭代自适应逆滤波器估计声门激励的方法,以声门激励的时域参数归一化振幅商作为特征,对六种不同情感的连续语音,首先使用F-ratio准则判别其对情感的区分能力,然后运用混合高斯模型对语音情感进行建模和识别。采用eNTERFACE’05情感语音数据库中的语音,比较了以整句NAQ值作为特征和以元音段的NAQ值作为特征,以及主观感知的情感识别结果。实验表明元音段的NAQ值是一种具有判别力的语音情感特征。This paper described a new time-domain parameter of the glottal flow, the normalized amplitude quotient and iteratire adaptive inverse filtering. Six emotional speech coming from the eNTERFACE＇05 audio-visual emotion database are inverse filtered using IAIF to estimate the glottal flow and parameterized using NAQ. To evaluate the properties of the emotion features based on the NAQ values of the vowel segments and of the whole speech sentences, firstly, gave F-ratio criterion to measure their ability to distinguish different emotions. Then, used the NAQ features to train the Gaussian mixture models for each speech emotional state, and compared emotion recognition results with subjective emotion perception results. Experimental results show that NAQ value of vowel segments can be used as an effective emotion feature in emotion recognition from speech.

关 键 词：迭代自适应逆滤波 归一化振幅商 F-ratio准则 混合高斯模型 

分 类 号：TP391.42[自动化与计算机技术—计算机应用技术]