Unified Framework for Single Channel Speech Enhancement

Ivan Tashev, Andrew Lovitt, and Alex Acero

Abstract

In this paper we describe a generic architecture for single channel speech enhancement. We assume processing in frequency domain and suppression based speech enhancement methods. The framework consists of a two stage voice activity detector, noise variance estimator, a suppression rule, and an uncertain presence of the speech signal modifier. The evaluation corpus is a synthetic mixture of a clean speech (TIMIT database) and in-car recorded noises. Using the framework multiple speech enhancement algorithms are tuned for maximum performance. We propose a formalized procedure for automated tuning of these algorithms. The optimization criterion is a weighted sum of the mean opinion score (PESQ-MOS), signalto-noise-ratio (SNR), log-spectral distance (LSD), and mean square error (MSE). The proposed framework provides a complete speech enhancement chain and can be used for evaluation and tuning of other suppression rules and voice activity detector algorithms.

Details

Publication typeInproceedings
Published in2009 IEEE Pacific Rim Conference on Communications, Computers and Signal Processing
AddressVictoria B.C., Canada
PublisherIEEE
> Publications > Unified Framework for Single Channel Speech Enhancement