An Alphanet approach to optimising input transformations for continuous speech recognition | IEEE Conference Publication | IEEE Xplore