See me, hear me: integrating automatic speech recognition and lip-reading