Comparison of Several Compensation Techniques for Robust Speaker Verification
It is well known that the performance of speaker recognition systems degrade rapidly as the mismatch between the training and test conditions increases. Thus, for example, in real-world telephone-based speaker recognition systems, both, additive and convolutional noise influence the error rate considerably. In this paper, different techniques which make a speaker verification system more robust against noise are described and compared. Some of these techniques have already been successfully applied in Robust Speech Recognition, and our preliminary results show that they are also very encouraging for Robust Speaker Verification.
Unable to display preview. Download preview PDF.
- C. García-Mateo and L. Rodriguez-Liñares. Speaker recognition based on a weighted acoustic discrimination. In EUSIPCO 96, volume III, pages 1047–1050, Trieste, September 1996.Google Scholar
- C. G.-M. L. Docío-Fernández. Application of several channel and noise compensation techniques for robust speaker recognition. In EUROSPEECH 97, September 1997.Google Scholar
- M. Rahim and B.-H. Juang. Signal bias removal by maximum likelihood estimation for robust telephone speech recognition. IEEE Trans. on ASP, 1 (4): 19–30, January 1996.Google Scholar
- D. Reynolds. Speaker identification and verification using gaussian mixture speaker models. In Proc. ESCA Workshop on Automatic Speaker Recognition, pages 27–l30, 1994.Google Scholar
- A. Sankar and C.-H. Lee. Robust speech recognition based on stochastic matching. In ICASSP 95, volume I, pages 125–124, 1995.Google Scholar