UWEETR-2004-0017 Author(s): Keywords: Abstract This paper is focused on the MVA (mean subtraction, variance normalization, and ARMA filtering) feature post-processing scheme for noise-robust automatic speech recognition. MVA has shown great success in the past on the Aurora 2.0 and 3.0 corpora. To test its generality, in this work MVA is blindly applied to many different acoustic feature extraction methods, and is evaluated using the Aurora 2.0 corpus. Specifically, we apply MVA post-processing to feature extraction techniques including: MFCC, LPC, PLP, RASTA, Tandem, Modulation-filtered Spectrogram and Modulation Cross-CorreloGram. We find that while effectiveness depends on the extraction method used, the majority of features benefit significantly from MVA. We conclude with a brief analysis. |