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:使用本征语音自适应的递归扬声器自适应自动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 recursively speaker adaptive speech recognition system and to a method, recognizes the input talker speech 1st, and compare with the negative results of the primary recognition four (Unsupervised Adaptation) method using a specific sound speaker adaptation own voice (EigenVoice) may be performed after the speaker adaptation, using the speaker-adapted acoustic models to further improve the recognition rate by recognizing the speaker's voice to the second enemy.; The invention of using the extracted the feature vectors used in speech recognition from an input speech signal, the voice characteristic extraction unit, and a speaker independent acoustic model has already been trained and the extracted feature vectors recognition vocabulary of the speech signals 1 and primary recognizer, the first recognition unit recognizes the result of the label information with the voice characteristic extraction unit feature vector in reference to observed data to the speaker adaptation section for performing speaker adaptation, and speaker dependent acoustic models provided by the speaker adaptation section used to be constituted by the second recognition and outputting a recognition result of the speech signal.
展开▼