Audio visual speech recognition based on multi-stream DBN models with Articulatory Features | IEEE Conference Publication | IEEE Xplore