Model-driven speech enhancement for multisource reverberant environment (Signal Separation Evaluation Campaign (SiSEC) 2011)

Abstract

We present a low complexity speech enhancement technique for real-life multi-source environments. Assuming that the speaker identity is known a priori, we present the idea of incorporating speaker model to enhance a target signal corrupted in non-stationary noise in a reverberant scenario. Based on experiments, this helps to improve the limited performance of noise-tracking based speech enhancement methods under unpredictable and non-stationary noise scenarios. Using pre-trained speaker model captures a constrained subspace for target speech and is capable to provide enhanced speech estimate by rejecting the non-stationary noise sources. Experimental results on Signal Separation Evaluation Campaign (SiSEC) showed that the proposed approach is successful in canceling the interference signal in the noisy input and providing an enhanced output signal. © 2012 Springer-Verlag.