摘要 |
PURPOSE: To provide a voice recognition device in which combined processing for the normalization and environmental adaptation of a spectrum is carried out. CONSTITUTION: An inputted voice is transmitted to an analysis section 10, in which the voice is analyzed into feature vectors, which are sent to a normalizing section 20. In the section 20, the feature vectors are normalized and the normalized feature vectors are transmitted to an adaptive section 30. In the section 30, the standard patterns sent from a standard pattern storage section 50 are made adaptive by means of the feature vectors sent from the section 20 and the standard patterns made adaptive in the section 30 are transmitted to a recognition section 40, in which voice recognition is conducted employing the feature vectors sent from the section 20 and the standard patterns sent from the section 30. |