Improved algorithms for speech recognition in noise using lateral inhibition and SNR weighting