METHOD OF MULTI MODEL ADAPTIVE AND VOICE RECOGNITION

摘要

PURPOSE: A multi model adaptive method and a voice recognition method are provided to enable a recognition system to use a different independent model according to each person or group during voice recognition adaptation. CONSTITUTION: A speaker selects one desired model number by a model number selecting unit(S101). If a voice of the speaker is inputted, a feature extracting unit extracts a feature vector necessary for adaptation from the voices(S102,S103). An adaptation processing unit performs adaptation for the feature vector by a set pronunciation information string model and a basic voice model(S104). If the adaptation is completed, the speaker stores a selected adaptation model by the adaptation processing unit(S105).