The report deals with speech enhancement in a noisy environment. A new, direct shaping parametric formulation of generalised adaptive spectral subtraction algorithm for non-stationary interference is considered. The idea of the authors is not to select but to form desirable envelope of spectral estimator function in SNR*Gain-function area. Some principles of parameters choice are proposed, based not only on aural estimations of signal quality but also on graphical analysis of target test signals. The field examinations showed good working performance of the proposed approach for noisy speech signal enhancement. The accuracy of various speech recognisers with the described pre-processing was significantly improved for noisy input speech. The paper develops approach originally reported in [5].
Cite as: Koval, S., Stolbov, M., Khitrov, M. (1999) Broadband noise cancellation systems: new approach to working performance optimization. Proc. 6th European Conference on Speech Communication and Technology (Eurospeech 1999), 2607-2610, doi: 10.21437/Eurospeech.1999-574
@inproceedings{koval99_eurospeech,
author={Serguei Koval and Mikhail Stolbov and Mikhail Khitrov},
title={{Broadband noise cancellation systems: new approach to working performance optimization}},
year=1999,
booktitle={Proc. 6th European Conference on Speech Communication and Technology (Eurospeech 1999)},
pages={2607--2610},
doi={10.21437/Eurospeech.1999-574}
}