Joint training of front-end and back-end deep neural networks for robust speech recognition | IEEE Conference Publication | IEEE Xplore