EUROSPEECH 2001 Scandinavia
7th European Conference on Speech Communication and Technology
2nd INTERSPEECH Event

Aalborg, Denmark
September 3-7, 2001

                 

A System for Text Dependent Speaker Verification - Field Trial Evaluation and Simulation Results

H. Schalk (1), Herbert Reininger (2), Stephan Euler (3)

(1) Universitšt Frankfurt, Germany; (2) ATIP, Germany; (3) Robert Bosch, Germany

In a speaker verification system, an identity claim is made by an unknown speaker. An utterance of this speaker and a model of the speaker whose identity is claimed is compared. If the model and the utterance match well the claim is accepted otherwise it is rejected. Thus, two classes of errors can occur in a speaker verification system: false acceptances and false rejections. The Equal Error Rate (EER) is often used as a performance measure of a verification system and results if the system parameters are adjusted in such a way that the two kinds of errors are equal. To model a speaker, reference utterances of this speaker are recorded in an enrollment session. Modelling itself can be done either in the signal domain using Dynamic Time Warping (DTW) or with a stochastic model of the speaker using Hidden Markov Models (HMM).

Full Paper

Bibliographic reference.  Schalk, H. / Reininger, Herbert / Euler, Stephan (2001): "A system for text dependent speaker verification - field trial evaluation and simulation results", In EUROSPEECH-2001, 783-786.