Comparing spectrum estimators in speaker verification under additive noise degradation
Författare
Summary, in English
Different short-term spectrum estimators for speaker verification under additive noise are considered. Conventionally, mel-frequency cepstral coefficients (MFCCs) are computed from discrete Fourier transform (DFT) spectra of windowed speech frames. Recently, linear prediction (LP) and its temporally weighted variants have been substituted as the spectrum analysis method in speech and speaker recognition. In this paper, 12 different short-term spectrum estimation methods are compared for speaker verification under additive noise contamination. Experimental results conducted on NIST 2002 SRE show that the spectrum estimation method has a large effect on recognition performance and stabilized weighted LP (SWLP) and minimum variance distortionless response (MVDR) methods yield approximately 7 % and 8 % relative improvements over the standard DFT method at -10 dB SNR level of factory and babble noises, respectively in terms of equal error rate (EER).
Avdelning/ar
- Matematisk statistik
- Statistical Signal Processing Group
Publiceringsår
2012
Språk
Engelska
Sidor
4769-4772
Publikation/Tidskrift/Serie
Acoustics, Speech and Signal Processing (ICASSP), 2012 IEEE International Conference on
Länkar
Dokumenttyp
Konferensbidrag
Förlag
IEEE - Institute of Electrical and Electronics Engineers Inc.
Ämne
- Probability Theory and Statistics
Nyckelord
- speaker verification
- spectrum estimation
Conference name
2012 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)
Conference date
2012-03-25 - 2012-03-30
Conference place
Kyoto, Japan
Status
Published
Forskningsgrupp
- Statistical Signal Processing
- Statistical Signal Processing Group
ISBN/ISSN/Övrigt
- ISSN: 1520-6149
- ISBN: 978-1-4673-0044-5 (online)
- ISBN: 978-1-4673-0045-2 (print)