首页>
外国专利>
Text-Independent Speaker Identification Using Telephone Speech
Text-Independent Speaker Identification Using Telephone Speech
展开▼
机译:使用电话语音的与文本无关的说话人识别
展开▼
页面导航
摘要
著录项
相似文献
摘要
The present invention relates to a statement independent speaker identification method using a voice signal.; Important applications of the telephone network of the speaker identified by the channel response by the limited bandwidth of the transmission line being displayed is non-linear distortion which occurs due to the severe performance degradation.; The present invention is the method of using the independent component analysis mapping the speech feature vector to a new feature space, assuming the cepstral vectors of the voice signal collected by the various environment as a linear combination of some of the features function, independent component analysis how to use a separate cepstral vectors in the learning and recognition are presented using.; The new vector region of the present invention are emphasized is characteristic information of the speech repeated, the channel distortion randomly appears is suppressed to improve the sense Chemistry tracking of the cepstrum space to appear in a more robust space for a channel environment change have.
展开▼