9P-C-1 エンターテインメント動画の感性検索を目的とした自己組織化マップによるシーン分析 : 音声トラック情報によるシーン分類(C会場 大学院生・学部学生 奨励賞セッション)

抄録

In the recent years, many audio and video clips are delivered from video sites on the Internet. But, generally, it is difficult to search interesting video clips on the sites, because the clips often are not titled accurately. In this study, we focus on the KANSEI-search method for entertainment video clips. In this article, we introduced analyzing method of entertainment video clips using self-organizing maps. We analyzed MANZAI video clips, and we classified emotional scenes using FFT ofsound-track, as the scenes of comedians' talking or audiences' laughing. As the result, using our proposal method, classification accuracy is 95 percent.