Rose, Siohan, 2022. End-to-end multi-talker audio-visual ASR using an active speaker attention module.