CONF lathoud05e/IDIAP Unsupervised Spectral Subtraction for Noise-Robust ASR Lathoud, Guillaume Magimai-Doss, Mathew Mesot, Bertrand Bourlard, Hervé EXTERNAL http://publications.idiap.ch/attachments/papers/2005/lathoud05e.pdf PUBLIC http://publications.idiap.ch/index.php/publications/showcite/lathoud-rr-05-42 Related documents Proceedings of the 2005 IEEE ASRU Workshop 2005 San Juan, Puerto Rico December 2005 IDIAP RR 05-42 This paper proposes a simple, computationally efficient 2-mixture model approach to discriminate between speech and background noise at the magnitude spectrogram level. It is directly derived from observations on real data, and can be used in a fully unsupervised manner, with the EM algorithm. In this paper, the 2-mixture model is used in an ``Unsupervised Spectral Subtraction'' scheme that can be applied as a pre-processing step for any acoustic feature extraction scheme, such as MFCCs or PLP. The goal is to improve noise-robustness of the acoustic features. Experimental results on both OGI Numbers 95 and Aurora 2 tasks yielded a major improvement on all noise conditions, while retaining a similar performance on clean conditions. REPORT lathoud-rr-05-42/IDIAP Unsupervised Spectral Substraction for Noise-Robust ASR Lathoud, Guillaume Magimai-Doss, Mathew Mesot, Bertrand Bourlard, Hervé EXTERNAL http://publications.idiap.ch/attachments/reports/2005/rr-05-42.pdf PUBLIC Idiap-RR-42-2005 2005 IDIAP Martigny, Switzerland Published in Proceedings of the 2005 IEEE ASRU Workshop This paper proposes a simple, computationally efficient \mbox{2-mixture} model approach to discriminate between speech and background noise at the magnitude spectrogram level. It is directly derived from observations on real data, and can be used in a fully unsupervised manner, with the EM algorithm. In this paper, the 2-mixture model is used in an ``Unsupervised Spectral Substraction'' scheme that can be applied as a pre-processing step for any acoustic feature extraction scheme, such as MFCCs or PLP. The goal is to improve noise-robustness of the acoustic features. Experimental results on both OGI~Numbers~95 and Aurora~2 tasks yielded a major improvement on all noise conditions, while retaining a similar performance on clean conditions.