4-8 映像コンテンツ特徴のパーソナライズ化とその応用(第4部門 ヒューマンインフォメーション)

抄録

In this paper, we propose to extract personalized video features based on the viewer's interest. The viewer's interest toward the video content is modeled in real time by monitoring their pupil size, gazing point and heart rate. The experimental results show that the video summary generated using weighted video features achieves higher precision and recall ratios for video with high narration nature if compared to video summary generated by video features without personalized weight prediction.