System and method for providing improved claimant authentication
View Patent ↗A system and method are provided for improving the accuracy of speaker authentication by combining the results of multiple verification sources using statistical modeling. A statistical classifier is built on top of the underlying verification sources. The classifier combines the multiple verification inputs with associated models of the respective verification technologies underlying the verification inputs. The classifier functions to classify a claimant as a true speaker or impostor based upon the verification inputs, the model or models and, potentially, additional information related to the context in which the classifier is being used.
1. An improved claimant authentication system, comprising:
a classifier configured to accept inputs from a plurality of verifier interfaces, wherein said plurality of verifier interfaces comprises at least one voice verifier and one knowledge verifier; and
a plurality of population model sets, each said population model set corresponding to a respective one of said plurality of verifier interfaces and comprising a first model of true speakers and a second model of impostors;
wherein said second model of imposters corresponding to said at least one knowledge verifier comprises a probability that an imposter knows personal identity context information associated with a true speaker of said first model;
wherein said classifier combines said inputs with said population model sets to provide a combined authentication result.
2. The system of claim 1 , further comprising:
a plurality of interface models corresponding to said plurality of verifier interfaces for providing said classifier with additional information relating to an accuracy of an underlying verification technology of said plurality of verifier interfaces, wherein said classifier combines said additional information with said inputs and said population model sets to provide a combined authentication result.
3. The system of claim 2 , further comprising:
a context information set for providing at least a first piece of information to said classifier relating to a content of said authentication system or component thereof, wherein said classifier combines said context information set with said inputs, said population model sets and said interface models to provide a combined authentication result.
4. The system of claim 1 , further comprising:
a context information set for providing at least a first piece of information to said classifier relating to a content of said authentication system or component thereof, wherein said classifier combines said context information set with said inputs and said population model sets to provide a combined authentication result.
5. An improved speaker authentication system, comprising:
a statistical classifier configured to accept a plurality of inputs from a corresponding plurality of verifier interfaces, wherein said plurality of verifier interfaces comprises at least one voice verifier and one knowledge verifier; and
a plurality of model sets coupled to said statistical classifier and operative to provide said statistical classifier with statistical models, wherein said statistical models are based upon information derived from an underlying verification technology corresponding to said verification interfaces;
wherein at least two of said plurality of model sets each comprise a first model of true speakers and a second model of impostors;
wherein said second model of imposters corresponding to said at least one knowledge verifier comprises a probability that an imposter knows personal identity context information associated with a true speaker of said first model;
wherein said statistical classifier combines said plurality of inputs with said plurality of statistical models to provide a combined authentication result.
6. The system of claim 5 , further comprising:
a context information set coupled to said statistical classifier for providing at least a first piece of information to said statistical classifier relating to a content of said authentication system or component thereof, wherein said statistical classifier combines said first piece of information with said inputs and said plurality of statistical models to provide a combined authentication result.
7. The system of claim 5 , wherein at least one of said verifier interfaces is a voice verifier and further wherein one of said plurality of model sets includes a voice verifier model based upon information derived from the underlying verification technology of said voice verifier.
8. The system of claim 7 , wherein said voice verifier model is constructed using a combination of interface information and population information.
9. The system of claim 8 , wherein said statistical classifier combines said plurality of inputs and said plurality of statistical models with context information to provide a combined authentication result.
10. The system of claim 9 , wherein said statistical classifier is a Bayes classifier.
11. A method for improving the accuracy of claimant authentication, comprising the steps of:
utilizing inputs from a plurality of verifier interfaces, wherein said plurality of verifier interfaces comprises at least one voice verifier and one knowledge verifier;
utilizing a plurality of models constructed from information relating to an underlying verification technology of said plurality of verifier interfaces; and
wherein at least two of said plurality of models each comprise a first model of true speakers and a second model of imposters; and
wherein said second model of imposters corresponding to said at least one knowledge verifier comprises a probability that an imposter knows personal identity context information associated with a true speaker of said first model;
combining said inputs with said plurality of models using a statistical classifier to provide a combined authentication result.
12. The method of claim 11 , further comprising the step of:
combining at least a first piece of information relating to a content of said authentication system or component thereof with said inputs and said plurality of models using said statistical classifier to provide said combined authentication result.
13. The method of claim 12 wherein a first of said verifier interfaces is a voice verifier and further wherein a corresponding first model from said plurality of models is a voice verifier model.