Title :
Utilizing Compressibility in Reconstructing Spectrographic Data, With Applications to Noise Robust ASR
Author :
Borgstrom, Bengt J. ; Alwan, Abeer
Author_Institution :
Dept. of Electr. Eng., Univ. of California, Los Angeles, CA
fDate :
5/1/2009 12:00:00 AM
Abstract :
In this letter, we propose a novel algorithm for reconstructing unreliable spectrographic data, a method applicable to missing feature-based automatic speech recognition (ASR). We provide quantitative analysis illustrating the high compressibility of spectrographic speech data. The existence of sparse representations for spectrographic data motivates the spectral reconstruction solution to be posed as an optimization problem minimizing the lscr1-norm. When applied to the Aurora-2 database, the proposed missing feature estimation algorithm is shown to provide significant improvements in recognition accuracy relative to the baseline MFCC system. Even without an oracle mask, performance approaches that of the ETSI advanced front end (AFE) , with less complexity.
Keywords :
data compression; speech coding; speech recognition; Aurora-2 database; automatic speech recognition; missing feature estimation algorithm; noise robust ASR; optimization problem; oracle mask; quantitative analysis; spectrographic data reconstruction; spectrographic speech data compressibility; Acoustic noise; Automatic speech recognition; Image reconstruction; Mel frequency cepstral coefficient; Noise robustness; Sampling methods; Spatial databases; Speech analysis; Telecommunication standards; Working environment noise; Compressibility; missing features; noise robust automatic speech recognition; spectral reconstruction;
Journal_Title :
Signal Processing Letters, IEEE
DOI :
10.1109/LSP.2009.2016452