摘要
针对情感变异的语音分类与识别研究的需要,给出了情感变异的语音语料库的构建流程和制作规范,设计并建立了情感变异的语音语料库。现已录制的语料包含自然、高兴、悲伤和愤怒四种情感类型,共1 600条语句。为了证明该情感变异的语音语料库的有效性,对其所有数据进行了听辨实验检测。结果表明,设计并构建的情感变异的语音语料库具有较高的实用价值,为情感变异的语音的分类与识别的研究奠定了夯实的基础。
This paper presents construction process and production standards of emotional variation speech corpora in research for the classification and identification of emotional variation speech,together with the design and structure of an elicited emotional variation speech corpora.Four types of speech emotions involving being natural,happy,sad and angry are recorded,a total of 1 600 authentic utterances.To demonstrate its validity,perception experiments on all the data have been conducted.The results show that the design and construction of the emotional variation speech corpora have a higher practical value for the classification and identification research of emotional variation speech.
作者
金玉卿
吕成国
JIN Yuqing,LV Chengguo(School of Computer Science and Technology,Heilongjiang University,Harbin 150080,China)
出处
《智能计算机与应用》
2011年第3期66-68,共3页
Intelligent Computer and Applications
基金
哈尔滨市青年科技创新人才基金资助项目(2007RFQXG097)
关键词
情感变异
语料库
分类
识别
听辨实验
Emotional Variation
Corpora
Classification
Identification
Perception Experiments