Multi-modal temporal asynchronicity modeling by product HMMs for robust audio-visual speech recognition | IEEE Conference Publication | IEEE Xplore