Deep Laterally Recurrent Spiking Neural Networks for Speech Enhancement
Julie A. Wall · UEL Research Repository (University of East London) · 2016
Compared to human performance ASR systems perform badly in noisy environments• Traditional signal processing approaches to noise reduction only partially successful • Lack of spatial information in the encoding of sound e.g.single channel GSM encoding • Is there some way we can make use of the rich spectro-temporal information to separate speech from noise?