A hybrid visual feature extraction method for audio-visual speech recognition | IEEE Conference Publication | IEEE Xplore