Feature mapping and recuperation by using elliptical basis function networks for robust speaker verification
- 13 Downloads
Abstract
The performance of speaker verification systems is often compromised under real-world environments. For example, variations in handset characteristics could cause severe performance degradation. This paper presents a novel method to overcome this problem by using a non-linear handset mapper. Under this method, a mapper is constructed by training an elliptical basis function network using distorted speech features as inputs and the corresponding clean features as the desired outputs. During feature recuperation, clean features are recovered by feeding the distorted features to the feature mapper. The recovered features are then presented to a speaker model as if they were derived from clean speech. Experimental evaluations based on 258 speakers of the TIMIT and NTIMIT corpuses suggest that the feature mappers improve the verification performance remarkably.
Key words
feature mapping and recurpuration elliptical basis function (EBF) networks speaker verificationPreview
Unable to display preview. Download preview PDF.
References
- [1]Tohkura Y. A weighted cepstral distance measure for speech recognition[J]. IEEE Transactions on Acoustics, Speech and Signal Processing, 1987,35(10): 1414–1422.CrossRefGoogle Scholar
- [2]Assaleh K T, Mammone R J. New LP-derived features for speaker identification [J]. IEEE Transactions on Speech and Audio Processing, 1994,2(4): 630–638.CrossRefGoogle Scholar
- [3]Atal B S. Effectiveness of linear prediction characteristics of the speech wave for automatic speaker identification and verification[J]. Journal of Acoustic Society of American, 1974,55(6): 1304–1312.CrossRefGoogle Scholar
- [4]Hermansky H, Morgan N. RASTA processing of speech [J]. IEEE Transactions on Speech and Audio Processing, 1994,2(4): 587–589.CrossRefGoogle Scholar
- [5]Rahim M G, Juang B H. Signal bias removal by maximum likelihood estimation for robust telephone speech recognition[J]. IEEE Transactions on Speech and Audio Processing, 1996,4(1): 19–30.CrossRefGoogle Scholar
- [6]Lo T F, Mak M W, Yiu K K. A new cepstrum-based channel compensation method for speaker verification[C]. Proc. Eurospeech ’99, Sept. 1999,2:775–778.Google Scholar
- [7]Mak M W, Kung S Y. Estimation of elliptical basis function parameters by the EM algorithm with application to speaker verification [J]. IEEE Transactions on Neural Networks, 2000,11 (4): 961–969.CrossRefGoogle Scholar
- [8]Jankowski C, Kalyanswamy A, Basson S, Spitz J. NTIMIT: A phonetically balanced, continuous speech, telephone bandwidth speech database [C]. Proc. ICAS-SP’ 90, 1990:109–112.Google Scholar
- [9]Zhang W D, Mak M W, Li C K, He M X. A priori threshold determination for phrase-prompted speaker verification[C]. Proc. Eurospeech ’99, 1999,2:1023–1026.Google Scholar
- [10]Zhang W D, Mak M W, He M X. A two-stage scoring method combining world and cohort model for speaker verification[C]. Proc. ICASSP’2000, June 2000.Google Scholar
- [11]Mak M W, Li C K, Li X. Maximum likelihood estimation of elliptical basis function parameters with application to speaker verification[C]. International Conference on Signal Processing (ICSP’98), Oct. 1998:1287–1289.Google Scholar
- [12]Li X, Mak M W. Optimization of EBFN architecture by an improved RPCL algorithm with application to speaker verification[C]. Proc. of the 1998 Symposium on Image, Speech, Signal Processing and Robotics (ISSPR ’98), 1998:2:43–47.Google Scholar