Noise reduction for randomized speech and audio coding in WASNs

We are surrounded by a multitude of connected devices with microphones, the signal of which should be combined for best sound quality. Thus, we recently proposed a distributed speech and audio codec that decorrelates quantization noise applying randomization. In this paper this method is extended attenuating quantization noise using Wiener filtering at the decoder. We demonstrate that this approach can be used to jointly attenuate quantization noise and background noise present at the microphones. By using orthogonal randomization matrices, computational complexity can be minimized by separating the Wiener ?lter from the inverse randomization. Our evaluation shows that Wiener filtering in combination with a randomized distributed codec is an efficient method to attenuate background and quantization noise at the decoder.