Combined speech enhancement and auditory modelling for robust distributed speech recognition