Multitaper Estimation of Frequency-Warped Cepstra with Application to Speaker Verification
Författare
Summary, in English
Usually the mel-frequency cepstral coefficients are estimated either from a periodogram or from a windowed periodogram. We state a general estimator which also includes multitaper estimators. We propose approximations of the variance and bias of the estimate of each coefficient. By using Monte Carlo computations, we demonstrate that the approximations are accurate. Using the proposed formulas, the peak matched multitaper estimator is shown to have low mean square error (squared bias + variance) on speech-like processes. It is also shown to perform slightly better in the NIST 2006 speaker verification task as compared to the Hamming window conventionally used in this context.
Avdelning/ar
- Matematisk statistik
- Statistical Signal Processing Group
Publiceringsår
2010
Språk
Engelska
Sidor
343-346
Publikation/Tidskrift/Serie
IEEE Signal Processing Letters
Volym
17
Issue
4
Dokumenttyp
Artikel i tidskrift
Förlag
IEEE - Institute of Electrical and Electronics Engineers Inc.
Ämne
- Probability Theory and Statistics
Nyckelord
- Speech analysis
- Multitapers
- Speaker verification
- Cepstral analysis
- Multiple windows
- MFCC
Status
Published
Forskningsgrupp
- Statistical Signal Processing Group
ISBN/ISSN/Övrigt
- ISSN: 1070-9908