Deep complementary bottleneck features for visual speech recognition.

Stavros Petridis, Maja Pantic

Browse the full ICASSP paper archive.