Goal detection in soccer video using audio/visual keywords

In this paper, we propose a two-level framework to detect interesting events automatically based on Audio and Visual Keywords (AVKs). The first level extracts low-level features such as motion. color, texture, pitch etc to detect video segments boundaries and label segments as audio and visual keywords. Next, we extract the exciting break portions from the AVK sequence. Then, we use two Hidden Markov Models (HMM) to model the exciting break portions with and without goal event respectively. We have applied the proposed approach to the detection of goal event in six half matches of soccer videos (270 minutes. 14 goals) from FIFA 2002 and UEFA 2002 and achieve 90% precision and 100% recall respectively.