US 6,205,424 B1Granted

Two-staged cohort selection for speaker verification system

AssigneeCompaq Computer Corporation

InventorsWilliam D. Goldenthal, Brian S. Eberman

Life of this patent

Estimating the term…

Loading patent…

Abstract

Speech signals from speakers having known identities are used to create sets of acoustic models. The acoustic models along with their corresponding identities are stored in a memory. A plurality of sets of cohort models that characterize the speech signals are selected from the stored sets of acoustic models, and linked to the set of acoustic models of each identified speaker. During a testing session speech signals produced by an unknown speaker having a claimed identity are processed to generate processed speech signals. The processed speech signals are compared to the set of models of the claimed speaker to produce first scores. The processed speech signals are also compared to the sets cohort models to produce second scores. A subset of scores are dynamically selected from the second scores according to a predetermined criteria. The unknown speaker is validated as the claimed speaker if the difference between the first and a combination of the subset of scores is greater than a predetermined threshold value.

Claims28

Dependent claims

A method for verifying a claimed identity of a speaker, the method comprising the steps of:

processing training speech signals to create sets of acoustic models, the training speech signals being produced by speakers having known identities during training sessions, there being one set of acoustic models for each identified speaker;
selecting from the sets of acoustic models a plurality of sets of cohort models for each identified speaker, each of the plurality of sets of cohort models characterizing the speech of a corresponding identified speaker;
processing testing speech signals produced by an unknown speaker claiming to be one of the identified speakers to generate processed testing speech signals during a testing session;
comparing the processed testing speech signals and the set of acoustic models for the claimed identified speaker to determine first scores;
comparing the processed testing speech signals and the plurality of sets of cohort models for the claimed identified speaker to determine second scores;
selecting a subset of scores from the second scores, the subset of scores corresponding to sets of cohort models which characterize the testing speech signals according to a predetermined criteria;
validating that the unknown speaker is the claimed identified speaker if the difference between the first scores and a predetermined combination of the subset of scores is greater than a predetermined threshold value.

4 more elements

Dependencies are inferred from the claim text.