Robust MVDR-based feature extraction for speech recognition

Author

Seyedin, Sanaz ; Ahadi, Seyed Mohammad

Author_Institution

Electr. Eng. Dept., Amirkabir Univ. of Technol., Tehran, Iran

fYear

2009

fDate

8-10 Dec. 2009

Firstpage

1

Lastpage

5

Abstract

This paper presents a novel noise robust feature extraction method for speech recognition. It is based on making the Minimum Variance Distortionless Response (MVDR) power spectrum estimation method robust against noise. This robustness is obtained by modifying the distortionless constraint of the MVDR spectral estimation method via weighting the subband power spectrum values based on the sub-band signal to noise ratios. The above method, when evaluated on Aurora 2 task, outperformed both the MFCC features as the baseline and the MVDR-based features in different noisy conditions.

Keywords

feature extraction; spectral analysis; speech recognition; MVDR spectral estimation; distortionless constraint; minimum variance distortionless response; power spectrum estimation; robust MVDR-based feature extraction; robust feature extraction; speech recognition; subband power spectrum values; Distortion; Feature extraction; Finite impulse response filter; Mel frequency cepstral coefficient; Noise robustness; Power generation; Signal to noise ratio; Spectral analysis; Speech recognition; Working environment noise; feature extraction; robust MVDR power spectral estimation; speech recognition;

fLanguage

English

Publisher

ieee

Conference_Titel

Information, Communications and Signal Processing, 2009. ICICS 2009. 7th International Conference on

Conference_Location

Macau

Print_ISBN

978-1-4244-4656-8

Electronic_ISBN

978-1-4244-4657-5

Type

conf

DOI

10.1109/ICICS.2009.5397503

Filename

5397503