A fast match for continuous speech recognition using allophonic models | IEEE Conference Publication | IEEE Xplore