Automatic recommendation system of IPTV contents for baseball video

In this paper, we propose a system to analyze baseball videos for creating video annotation and recommendation. For video annotation, we treat the video segment between two pitch shots as a event. The caption is inferred to find categories of the event. A rule-based decision tree is used to classify a event into four event categories. For video recommendation, users are recommended items that according to highlight events. They can browse their favorite events in a baseball video. The algorithm is tested on 68 events from US and Taiwan baseball video. All events are rated in a 1-to-5 rating scale. The highest average rating is 4.4.