OCR-aided person annotation and label propagation for speaker modeling in TV shows.

Mateusz Budnik, Laurent Besacier, Ali Khodabakhsh, Cenk Demiroglu

Browse the full ICASSP paper archive.