情感语音合成技术或对声纹鉴定准确性产生影响  

Effect of Emotional Speech Synthesis Techniques on the Accuracy of Voiceprint Identification

作　　者：白海莉 

机构地区：[1]甘肃政法学院公安技术学院,甘肃兰州730070

出　　处：《科技创新与应用》2018年第36期24-24,26,共2页Technology Innovation and Application

摘　　要：声纹鉴定运用语言学、计算机科学等知识,对涉案语音和样本语音进行听觉辨识,同时运用频谱图进行综合分析,判断二者是否由同一音源发出,或者判断声音性质。声纹鉴定结果可以为侦查和诉讼活动提供线索或证据。深度学习可以促进情感语音合成技术的发展,而情感语音合成技术的发展水平越高,合成语音与人声的差异就越小,声纹鉴定的难度就越大。文章旨在研究深度学习以及情感语音合成技术发展的发展状况,以此分析其可能对声纹鉴定准确性产生影响的因素。Voiceprint identification uses the knowledge of linguistics and computer science to identify the involved speech and the sample speech. At the same time, it uses the spectral analysis to determine whether the two are generated by the same sound source, or the nature of the voice. The results of voiceprint identification can provide clues or evidence for investigation and litigation activities. Deep learning can promote the development of emotional speech synthesis technology, and the higher the developmentlevel of emotional speech synthesis technology, the smaller the difference between synthetic speech and human voice, and the moredifficult it is to identify voiceprint. The purpose of this paper is to study the development of deep learning and emotional speech synthesis technology, and to analyze the factors that may affect the accuracy of voiceprint identification.

关 键 词：语音 声纹鉴定 深度学习 语音合成 情感语音合成 

分 类 号：TN912.33[电子电信—通信与信息系统]