Abstract: Recently, audio-visual speech recognition has attracted increasing attention. However, most existing works only focused on scenarios with two speakers. In this work, we study the effect of ...