摘要
提出了基于小波变换理论的说话人识别算法———小波基包法 .利用小波变换多分辨的特性 ,提取出真正地反映基音周期变化的基频包络 ,从声学语音学的角度去认识方言和口音 ,首次提出了汉语口音的参量表征 ,并应用到说话人识别当中 ,以实现从听觉的角度出发 ,进行说话人的先天特性与后天特性相结合的说话人识别 .结果表明 ,从基频包络中可以表现口音的一些表征 ,有助于认识说话人的特性 .
A speaker recognition, a pitch contour detector by wavelet transform, is presented. The evaluation of pitch periods is exactly performed by a pitch detector which is based on multiresolution of wavelet transform. In order to represent the dialect and the accent from the acoustic phonetics perspective, the Chinese accent is quantified for the first time, which is also applied to speaker identification to realize the recognition with the combination of the speaker′s innate and acquired characteristics.
出处
《西安电子科技大学学报》
EI
CAS
CSCD
北大核心
2000年第4期437-441,共5页
Journal of Xidian University
关键词
小波变换
说话人识别
口音特性
语音识别
wavelet transform
pitch contour
speaker recognition
accent characteristic
