Speaker identification system using word recognition templates

A method for recognizing the identity of a speaker comprising the steps of storing a set of templates each representative of the acoustic features of a reference word for a distinct group of speakers;

• generating for each of a plurality of identified speakers a set of first signals each representative of the correspondence of the identified speaker's utterance of said reference words with said templates for said reference words;
• generating a set of signals representative of the acoustic features of an utterance by an unknown speaker;
• recognizing the words of the utterance of said unknown speaker from the utterance feature signals and said templates;
• generating a set of second signals representative of the correspondence of the feature signals of the recognized words of the unknown speaker to the feature signals of said templates;
• comparing said first correspondence signals for the recognized words of each identified speaker to the second correspondence signals to identify the unknown speaker as the identified speaker having the closest matching correspondence signals.

Apparatus for recognizing a speaker comprising;

• means for storing a set of templates representative of the acoustic features of predetermined reference words;
• means responsive to the identified speaker's utterance and said templates for generating a set of first signals representative of the correspondence of the identified speaker's utterance of said reference words with said templates for each of a plurality of identified speakers;
• means for generating a set of signals representative of the acoustic features of the utterance of an unknown speaker;
• means responsive to the unknown speaker's utterance feature signals and said templates for recognizing the words of the unknown speaker's utterance;
• means responsive to the feature signals of the recognized words of the unknown speaker and the feature signals of said templates for generating a set of second signals representative of the correspondence of the recognized words of the unknown speaker with the templates for the recognized words;
• and means responsive to the first signals for the recognized words of the identified speakers and the second signals for identifying the unknown speaker.