An Effective Audio-Visual Information Based Framework for Extracting Highlights in Basketball Games

摘要

Recently, extracting highlights in the sports videos has been paid much attention. This paper proposes a hierarchical extracting model to extract the highlights in the basketball games based on the audio-visual features. Firstly, the unrelated scenes (studio scenes, Ad scenes) are removed by detecting music sound and analyzing scene background's stableness. Then, the clips which the audiences pay more attention to are marked by detecting the keywords and exciting sounds in the audio clues and the score changing in the visual clues. We do a series of experiments to evaluate the proposed method, and the experimental results show that our method can work well.