THRESHOLD SETTING AND TRAINING OF A SPEAKER VERIFICATION SYSTEM

摘要

Speaker verification system in comprising model building means which build n different speaker models, each based on n-1 tokens, for each model one independent token being available for estimating the distance between the model and an utterance that has not be used to build the model. Estimation means estimate the central tendency of the distance between the speaker's model and newly prodced utterances of the same speaker, and also its variance. The central tendency of distances between the model and utterances is optimised, based on linear interpolation: Th(new) = b * CTi + (1 -b) * CTt, where Th(new) is the optimal threshold, CTi is the central tendency obtained from pre-recorded impostor speech, CTt is the central tendency estimated from the enrolment speech of the new customer, and b is the interpolation parameter, that is optimised using additional pre-recorded utterances not used in estimating CTi.