Analysis of sidelobe blanking technique for two-channel speech enhancement

Abstract

This paper analyzes the application of Sidelobe Blanking Logic to Two-Channel speech enhancement. We show that several separately proposed Two-Channel post-filtering speech enhancement methods can be viewed as variants of Sidelobe Blanking Logic technique which first arouse in the Radar community around the 1970s. We show that the core mechanism of this kind of technique lies in the two combined target detection measures, that is nonstationarity and Main to Auxiliary ratio. Consequently, the key role played by the detection thresholds is revealed. From this point of view, we show that a well-known two-channel post-filtering method can be improved by adapting the threshold to the main and auxiliary receiver characteristics, and simplified by using a single hard threshold and wiener filtering instead of double thresholds and OM-LSA, without significant performance loss.