Interestingness Detection in Sports Audio Broadcasts

This paper presents a novel method for semantic understanding of sports matches by extracting and ranking events within a match by interestingness. Using audio feature extraction, a system is presented which is able to segment between studio and pitch side broadcast. Key events within Rugby Union matches are then identified based on crowd excitation levels and referee whistles. This identifies individual interesting events and a timeline of interestingness estimation allowing viewers to navigate to sections of the broadcast where interesting sections of play occur.