Speech discrimination in adverse conditions using acoustic knowledge and selectively trained neural networks | IEEE Conference Publication | IEEE Xplore