摘要
提出一种新型声纹识别系统的设计方案。有别于传统声纹识别系统的是,该识别系统的预处理部分使用言语过滤的方法,特征提取过程采用情感补偿的办法。其中,言语过滤以语音识别为技术支持,可以剔除嘈杂背景,限制冒名顶替,排除语音赝本等等;情感补偿以语音情感计算为理论基础,能够很好地解决声音漂移、伪装模仿等问题。基于这两部分提取特征数据将更加准确有效。
This paper proposes a new method of voiceprint recognition. Unlike traditional voiceprint recognition, speech filter is applied to preprocessing of the recognition system, and emotional compensation approach is used throughout the whole process of characteristics extraction. Among them, voice recognition is the technology support of speech filter. It can remove background noise, restrict impostors, and eliminate the counterfeiting of voice, etc. Emotional Compensation is based on the voice emotional calculation. This part can easily resolve some problems, such as voice drift, the camouflage and mimicry. The data extracted from the two parts are more accurate and efficient.
出处
《重庆科技学院学报（自然科学版）》
CAS
2007年第1期91-93,共3页
Journal of Chongqing University of Science and Technology：Natural Sciences Edition
关键词
声纹识别
言语过滤
活体声纹
情感补偿
voiceprint recognition
speech filter
live voiceprint
emotional compensation
