CAJ | 학술논문

在收集较大规模情感语音语料库基础上，分析了维吾尔语语音在韵律特征和音质特征方面的19种语境信息和6种情感特征参数，并利用STRAIGHT算法实现了情感特征参数的提取，最后利用分类回归树( CART)算法针对各个情感特征和中性向其它情感的转换特征进行了建模。实验结果表明，所提取情感特征能准确的区分各个情感类型，为实现中性语音转换成各种目标情感语音奠定了基础。
재수집교대규모정감어음어료고기출상，분석료유오이어어음재운률특정화음질특정방면적19충어경신식화6충정감특정삼수，병이용STRAIGHT산법실현료정감특정삼수적제취，최후이용분류회귀수( CART)산법침대각개정감특정화중성향기타정감적전환특정진행료건모。실험결과표명，소제취정감특정능준학적구분각개정감류형，위실현중성어음전환성각충목표정감어음전정료기출。
Based on large-scale emotional Uyghur speech database, this paper analyzes 19 contextual infor-mations and 6 emotional prosodic features of the speech data. STRAIGHT algorithm is used to implement the extraction of emotional characteristic parameters, and finally the classification and regression tree ( CART) algorithm is used to build up the individual models for each kind of emotional speech and models for emo-tional conversion from neutral to another given emotion. Experimental results show that the extracted emo-tional characteristics could be used to accurately distinguish various types of emotions. All this would lay a foundation for achieving the conversion of neutral speeches to various target emotional speeches.