SUBWORD-BASED SPEAKER VERIFICATION USING MULTIPLE CLASSIFIER FUSION, WITH CHANNEL, FUSION, MODEL, AND THRESHOLD ADAPTATION

摘要

The subword-based, text-dependent automatic speaker verification voiceprint system embodies a capability of user-selectable passwords with no constraints on the choice of vocabulary words or the language. Automatic blind speech segmentation allows speech to be segmented into subword units (210) without any linguistic knowledge of the password. Subword modeling is performed using multiple classifiers (240, 250). The system also takes advantage of such concepts as multiple classifier fusion (260) and data resampling to successfully boost the performance. Key word/key phrase spotting (200) is used to optimally locate the password phrase. Numerous adaptation techniques increase the flexibility of the base system, and include: channel adaptation (180), fusion adaptation (290), model adaptation (220, 230) and threshold adaptation (295).