Digital video volume has been increasing exponentially for the past decade in both professional and amateur domains. In order to skim through video archives rapidly and efficiently, it is crucial to have a system to automatically generate video summary. In this paper, we propose to generate human-interest-aware video summary by reflecting human interest to the low-level video features. Human interest is obtained by modeling human bio-signals while they are watching the video. A subjective feedback experiment is carried out, and the video summaries which take human interest into account are evaluated better in conveying the story, impression, importance and interest of the full video.