There has been a concerted effort from the Video Retrieval community to develop tools that automate the annotation process of Sports video. In this paper, we provide an in-depth investigation into three Hidden Markov Model (HMM) selection approaches. Where HMM, a popular indexing framework, is often applied in a ad hoc manner. We investigate what effect, if any, poor HMM selection can have on future indexing performance when classifying specific audio content. Audio is a rich source of information that can provide an effective alternative to high dimensional visual or motion based features. As a case study, we also illustrate how a superior HMM framework optimised using a Bayesian HMM selection strategy, can both segment and then classify Soccer video, yielding promising results.
|Title of host publication||Lecture Notes in Computer Science|
|Publication status||Published - 2004|
- information retrieval
- video retrieval
- learning objects
- hidden Markov model