首页>
外国专利>
Recursive Speaker Adaptation Automation Speech Recognition System and Method using EigenVoice Speaker Adaptation
Recursive Speaker Adaptation Automation Speech Recognition System and Method using EigenVoice Speaker Adaptation
展开▼
机译:使用EigenVoice说话人自适应的递归说话人自适应自动化语音识别系统和方法
展开▼
页面导航
摘要
著录项
相似文献
摘要
The present invention relates to an adaptive recursive speaker voice recognition system and method using a unique sound speaker adaptation, a first input audio speaker recognition, and after the first recognized using a negative result unsupervised (Unsupervised Adaptation) performing a specific voice (EigenVoice) speaker adaptation in the way, the speaker using the speaker adaptive acoustic model a speech recognition rate can be further improved by recognizing a second enemy. ; This invention provides a speech feature extractor for extracting a feature vector used for speech recognition from an input speech signal, the extracted and the first recognition unit recognizes the vocabulary of the speech signal by using the speaker-independent acoustic model has already been trained with the feature vectors, said first recognition portion recognizes the result of the label information and the speech characteristic extraction unit in the observation feature vector data It is referred to by using a speaker dependent acoustic model that is provided by the speaker adaptation unit and the speaker adaptation section for performing speaker adaptation constituted by the second recognition and outputting a recognition result of the speech signal.
展开▼