Noise suppression for speech enhancement is desirable in various speech applications. In this paper we present a solution to the nonlinear spectral estimation problem for speech enhancement. We start from rather simple statistical models for the short time spectral estimates of speech and noise (log-uniform or log-normal). By empirical data generation and a curve fitting approach we are able to get explicit, though simple, expressions for the MMSE estimator in function of input level and the model parameters for each frequency component. The great advantage of our approach is that it has a sound theoretical foundation, is general by the choice of its parameters, and almost as simple to use as classical spectral subtraction.
Keywords: MMSE, Spectral estimation
Cite as: Xie, F., Compernolle, D.V. (1993) Speech enhancement by nonlinear spectral estimation - a unifying approach. Proc. 3rd European Conference on Speech Communication and Technology (Eurospeech 1993), 617-620, doi: 10.21437/Eurospeech.1993-150
@inproceedings{xie93_eurospeech,
author={Fei Xie and Dirk Van Compernolle},
title={{Speech enhancement by nonlinear spectral estimation - a unifying approach}},
year=1993,
booktitle={Proc. 3rd European Conference on Speech Communication and Technology (Eurospeech 1993)},
pages={617--620},
doi={10.21437/Eurospeech.1993-150}
}