On the basis of the analytical review of the basic modern biometric technologies of the entry control, as well as methods and concrete systems of speaker verification, the discriminant-stochastic approach to construction of speech verifiers (identifiers) possessing high reliability characteristics of speaker recognition. The concrete structure of the discriminant-stochastic speaker verifier (DSSV) in which three types of models (integrated, VQ, and GMM) are applied as template of the client. They allow to use more widely a range of modular spectrum (from 0,125 up to 10 Hz) practical to verification and identification. The final decision on recognition in DSSV is adopted under three individual decisions independent from each other by voting. On the DSSV base the high-level systems of the authorized access to the information or physical access on objects with high danger can be created.
Cite as: Rylov, A.S., Chyzhdzenka, V.A., Leukouskaya, T.V. (2004) The discriminant-stochastic approach of the speaker verification for entry control by the biometrical technologies. Proc. 9th Conference on Speech and Computer (SPECOM 2004), 377-381
@inproceedings{rylov04_specom,
author={A. S. Rylov and V. A. Chyzhdzenka and T. V. Leukouskaya},
title={{The discriminant-stochastic approach of the speaker verification for entry control by the biometrical technologies}},
year=2004,
booktitle={Proc. 9th Conference on Speech and Computer (SPECOM 2004)},
pages={377--381}
}