Text-Prompted Multistep Speaker Verification Using Gibbs-Distribution-Based Extended Bayesian Inference for Reducing Verification Errors
This paper presents a method of text-prompted multistep speaker verification for reducing verification errors. The method is developed for our speech processing system which utilizes competitive associative nets (CAN2s) for learning piecewise linear approximation of nonlinear speech signal to extract feature vectors of pole distribution from piecewise linear coefficients reflecting nonlinear and time-varying vocal tract of the speaker. This paper focuses on reducing verification errors by means of multistep verification using Gibbs-distribution-based extended Bayesian inference (GEBI) in text-prompted speaker verification. The effectiveness of GEBI and the comparison to BI (Bayesian inference) is shown and analyzed by means of experiments using real speech signals.
KeywordsText-prompted multistep speaker verification Gibbs-distribution-based extended Bayesian inference Competitive associative net
Unable to display preview. Download preview PDF.
- 1.Beigi, H.: Fundamentals of speaker recognition. Springer-Verlag New York Inc. (C) (2011)Google Scholar
- 2.Melin, H., Lindberg, J.: Prompting of Passwords in Speaker Verification Systems, Fonetik-97, Phonum 4, Umeå University, Sweden, May 28-30 (1997)Google Scholar
- 3.Kurogi, S., Ueno, T., Sawa, M.: A batch learning method for competitive associative net and its application to function approximation. In: Proc. SCI 2004, vol. V, pp. 24–28 (2004)Google Scholar
- 5.Mizobe, Y., Kurogi, S., Tsukazaki, T., Nishida, T.: Multistep speaker identification using Gibbs-distribution-based extended Bayesian inference for rejecting unregistered speaker. In: Huang, T., Zeng, Z., Li, C., Leung, C.S. (eds.) ICONIP 2012, Part V. LNCS, vol. 7667, pp. 247–255. Springer, Heidelberg (2012)CrossRefGoogle Scholar