Discriminatively trained features using fMPE for multi-stream audio-visual speech recognition.

Jing Huang, Daniel Povey

Browse the full Interspeech paper archive.