摘要:
The present invention provides a system and method for performing speaker verification on a spoken utterance. The present invention includes a verification template generator (40) which generates speaker-dependent verification templates for each possible spoken item that may be in the spoken utterance. A speech recognizer (42) performs speech recognition functions on the spoken utterance. A frame mapper (44) marks the beginning and ending of each spoken item within the spoken utterance. Verification module (46) generates a verification signal in response to verifying a minimum number of the spoken items within the spoken utterance being matched with the speaker dependent verification templates.