Abstract

A time-varying Wiener filter extracts the speech signal from a noisy mixture using the a priori signal-to-noise ratio in a local time-frequency unit. We estimate this ratio using a binaural processor and derive a ratio time-frequency mask. This mask is used to extract the speech signal, which is then fed to a conventional speech recognizer operating in the cepstral domain. We compare the performance of this system with a missing data recognizer that operates in the spectral domain using the timefrequency units dominated by speech. For use by the missing data recognizer, the same processor is used to estimate an ideal time-frequency binary mask, which selects the speech signal if it is stronger than the interference in a local time-frequency unit. We find that the performance of the missing data recognizer is better on a small vocabulary recognition task but the performance of the conventional recognizer is substantially better when the vocabulary size is larger. 1.

Citations

1180	Fundamentals of speech recognition - Rabiner - 1993
776	DB: Statistical analysis with missing data - RJA, Rubin - 1987
550	Comparison of parametric representations of monosyllabic word recognition in continuously spoken sentences - Davis, Mermelstein - 1980
300	Suppression of Acoustic Noise in Speech using Spectral Subtraction - Boll - 1979
270	signal separation: statistical principles - Cardoso, “Blind - 1998
246	Speech enhancement using a minimum mean-square error short-time spectral amplitude estimator - Ephraïm, Malah - 1984
224	Spatial hearing: The psychophysics of human sound localization - Blauert - 1983
159	Robust automatic speech recognition with missing and unreliable acoustic data - Cooke, Green, et al. - 2001
132	Hidden Markov model decomposition of speech and noise - Varga, Moore - 1990
130	A database for speaker-independent digit recognition - Leonard - 1984
101	Speech recognition by machines and humans - Lippmann - 1997
100	The darpa 1000-word resource management database for contin 101 uous speech recognition - Price, Fisher, et al. - 1988
86	DARPA TIMIT acoustic-phonetic continuous speech corpus CD-ROM - Garofolo, Lamel, et al. - 1993
84	Detection, Estimation, and Modulation Theory Part I - Trees - 1968
81	Speech recognition in noisy environments: A survey,” Speech communication - Gong - 1995
78	Robust continuous speech recognition using parallel model combination - Gales, Young - 1996
78	Speech enhancement using a soft-decision noise suppression filter - McAulay, Malpass - 1980
71	The NOISEX-92 study on the effect of additive noise on automatic speech recognition - Varga, Steeneken, et al. - 1992
67	Separation of speech from interfering sounds based on oscillatorycorrelation. Neural Networks - Wang, Brown - 1999
65	New telephone speech corpora at cslu - Cole, Noel, et al. - 1995
61	Speech segregation based on sound localization - Wang, Brown - 2003
59	Soft decisions in missing data techniques for robust automatic speech recognition - Barker, Josifovski, et al. - 2000
52	HRTF measurements of a KEMAR dummyhead microphone - Gardner, Martin - 1994
37	Bayesian estimation approach for speech enhancement using hidden markov models - Ephraim, “A - 1992
37	Reconstruction of missing features for robust speech recognition - Raj, Seltzer, et al.
25	A binaural processor for missing data speech recognition in the presence of noise and small-room reverberation - Palomäki, Brown, et al.
15	Missing feature theory in ASR: make sure you missing the right type of features - Veth, Wet, et al. - 1999
15	Computational Auditory Scene Analysis. Lawrence Erlbaum Associates - Rosenthal, Okuno - 1998
14	A casa-labelling model using the localisation cue for robust cocktail-party speech recognition - Glotin, Berthommier, et al. - 1999
13	Reconstruction of damaged spectrographic features for robust speech recognition - Raj, Seltzer, et al. - 2000
13	Robust Speech Recognition Using Missing Feature Theory - hamme - 2003
12	Separation of speech by computational auditory scene analysis - Brown, Wang - 2005
12	Evaluation of a noiserobust DSR front-end on Aurora databases - Macho, Mauuary, et al. - 2002
10	Performance of an hmm speech recognizer using a real-time tracking microphone array as input - Hughes, Kim, et al. - 1999
8	Discriminant Training of Front-End and Acoustic Modeling Stages to Heterogeneous Acoustic Environments for Multi-stream Automatic Speech Recognition - Shire - 2000
7	A casa front-end using the localisation cue for segregation and then cocktail-party speech recognition - Tessier, Berthommier, et al. - 1999
7	A nonlinear observation model for removing noise from corrupted speech log mel-spectral energies - Droppo, Acero, et al. - 2002
7	Blind separation of convolutive mixtures and an application in automatic speech recognition in noisy environment - Ehlers, Schuster - 1997
6	Speech feature smoothing for robust ASR - Chen, Bilmes, et al. - 2005
6	The role of evidence and counter-evidence in speech perception - Cunningham, Cooke - 1999
3	Speech processing, transmission and quality aspects (STQ); distributed speech recognition; advanced front-end feature extraction algorithm; compression algorithms - STQ-AURORA
1	Speech Communication 48 (2006) 1486–1501 1501 - Srinivasan - 1999