Efficient real-time noise estimation without explicit speech, non-speech detection: an assessment on the AURORA corpus

N.W.D. Evans1, J.S. Mason1, B. Fauve2
1Department of Electronic and Electrical Engineering, University of Wales, Swansea, UK
2ENSPM, Marseilles, France

Tóm tắt

This paper addresses the problem of noise estimation for speech enhancement and automatic speech recognition. In the context of mobile telephony, there is a requirement for low resource algorithms which must run at real-time. This paper describes the implementation of a previously published approach, termed quantile-based noise estimation, integrated within a conventional spectral subtraction framework. The novelty lies in the efficiency of the noise estimation process. Assessment is carried out on the AURORA corpus and demonstrates significant improvements in efficiency. Automatic speech recognition results show an average relative improvement of 26% over the baseline.

Từ khóa

#Speech enhancement #Automatic speech recognition #Telephony #Speech recognition #Proposals #Noise robustness #Additive noise #Background noise #Network servers #Telecommunication traffic

Tài liệu tham khảo

cho, 2001, mixed decision-based noise adaptation for speech enhancement, Electronics Letters, 37, 540, 10.1049/el:20010368 droppo, 2001, Evaluation of the SPLICE Algorithm on the Auroral Database, Proc EUROSPEECH, 1, 217 hirsch, 2000, The Aurora Experimental Framework for the Performance Evaluation of Speech Recognition Systems Under Noisy Conditions, Proc ISCA ITRW ASR Automatic Speech Recognition Challenges for the Next Millennium martin, 1994, Spectral Subtraction Based on Minimum Statistics, Proc EUSlPCO, 1182 doblinger, 1995, Computationally Efficient Speech Enhancement by Spectral Minima Tracking in Subbands, Proc EUROSPEECH, 2, 1513 10.1109/ICASSP.1995.479387 evans, 2001, Noise Estimation Without Explicit Speech, Non-speech Detection: a Comparison of Mean, Median and Modal Based Approaches, Proc EUROSPEECH, 2, 893 press, 1999, Numerical Recipes in C ealey, 2001, Harmonic Tunnelling: Tracking Non-stationary Noises During Speech, Proc EUROSPEECH, 1, 437 yapanel, 2001, Robust Digit Recognition in Noise: An Evalutaion using the AURORA Corpus, Proc EUROSPEECH, 1, 209 arslan, 1995, New Methods. for Adaptice Noise Suppression, Proc ICASSP, 1, 812 lockwood, 1991, Experiments with a Nonlinear Spectral Subtractor (NSS), Hidden Markov Mooels and the Projection, for Robust Speech Recognition in Cars, Proc EUROSPEECH, 1, 79 10.1016/0885-2308(89)90027-2 10.1109/ICASSP.2000.862122 gales, 1993, HMM Recognition in Noise using Parallel Model Combination, Proc EU-ROSPEECH, 837 varga, 1991, Simultaneous Recognition of Concurrent Speech Signals using Hidden Markov Model Decomposition, Proc EUROSPEECH, 3, 1175 10.1109/ICASSP.1979.1170788 10.1109/TASSP.1979.1163209 attias, 2001, A New Method for Speech Denoising and Robust Speech Recognition using Probabalistic Models for Clean Speech and for Noise, Proc EUROSPEECH, 3, 1903 barker, 2001, Robust ASR Based On Clean Speech Models: An Evaluation of Missing Data Techniques For Connected Digit Recognition in Noise, Proc EUROSPEECH, 1, 213