UDC 534.78, 621.391.266
1 V.M. Glushkov Institute of Cybernetics, National Academy of Sciences of Ukraine, Kyiv, Ukraine
zvk140@ukr.net
|
|
3 Institute of Mathematics, National Academy of Sciences of Ukraine, Kyiv, Ukraine, and Kyiv Academic University, Kyiv, Ukraine
semenovaevgen@gmail.com
|
|
METHOD OF NOISE-ROBUST ESTIMATION OF PARAMETERS
OF AUTOREGRESSIVE MODEL IN FREQUENCY DOMAIN
Abstract. The article considers the problem of estimating the parameters of the autoregressive (AR) signal in the presence of background noise.
Based on the frequency representation of the AR signal, a technique of calculating the likelihood function
of the AR parameters is shown and the implementation of the expectation-maximization method
for iterative evaluation of the AR parameters is considered. Analysis of different measures of distortion
of speech signals shows that the proposed approaches in the frequency domain have the same accuracy
with the corresponding approaches in the time domain, but are characterized by significantly lower computing costs.
Keywords: autoregressive model, likelihood function, Expectation-Maximization method, fast Fourier transform.
FULL TEXT
REFERENCES
- Markel J.D., Gray A.H. Linear prediction of speech [Russian translation]. Moscow: Radio i svyaz', 1980. 307 p.
- Dempster A., Lair N., Rubin D. Maximum likelihood from incomplete data via the EM algorithm. J. Roy. Statistic. Soc. 1977. Vol. 39. P. 1–38.
- Gannot S., Burnstein D., Weinstein E. Iterative and sequential Kalman filter-based speech enhancement algorithms. Transactions Speech Audio Processing. 1998. Vol. 6. P. 373–385.
- Voevodin V.V., Tyrtyshnikov E.E. Computational processes with Toeplitz matrices [in Russian]. Moscow: Nauka, 1987. 320 p.
- Sage A., Melsa J. Estimation theory and its application in communication and control. Series: Statistical Communication [Russian translation] Theory. Moscow: Radio i svyaz', 1976. 496 p.
- Semenov V.Yu. Methods for calculating and coding the parameters of an autoregressive speech model when developing a vocoder based on a fixed-point signal processor. Problemy upravleniya i informatiki. 2019. N 1. P. 41–50.
- Zadiraka V.K., Melnikova S.S. Digital signal processing [in Russian]. Kyiv: Nauk. dumka, 1993. 294 p.
- Zadiraka V.K. Fourier transform computation theory [in Russian]. Kyiv: Nauk. dumka, 1983. 216 p.
- Lim J.S., Oppenheim A.V. All–pole modeling of degraded speech. IEEE Trans. Acoust. Speech Signal Proces. 1978. Vol. 26. P. 197–210.