Presentation 2004/6/11
Video Indexing for Human Behavior Analysis by Robust Face and Hands Detection
Hitoshi IKEDA, Masahiro MAEDA, Noriji KATO, Hirotsugu KASHIMURA,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) In this paper, we describe a novel classification technique that separates video file into several scenes according to human actions. We use face coordinates, angle in depth, and hands coordinates from a video frame as an input for classification. To detect face coordinates and angle accurately, we introduce a latent variable model in kernel-induced feature space. The model codes variations of face image robustly in manner of the population coding. To detect hands coordinates reliably, we introduce color-based method that searches face color area as hand area. For classification, we use the maharanobis distance of the input to evaluate similarity of video frames, and then we use the hierarchical clustering method to bind video frames. We apply this classification technique to separate office work video into work scenes.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Video / Indexing / Face Detection / Hands Detection / hierarchical clustering
Paper # PRMU2004-43
Date of Issue

Conference Information
Committee PRMU
Conference Date 2004/6/11(1days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair
Vice Chair
Secretary
Assistant

Paper Information
Registration To Pattern Recognition and Media Understanding (PRMU)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Video Indexing for Human Behavior Analysis by Robust Face and Hands Detection
Sub Title (in English)
Keyword(1) Video
Keyword(2) Indexing
Keyword(3) Face Detection
Keyword(4) Hands Detection
Keyword(5) hierarchical clustering
1st Author's Name Hitoshi IKEDA
1st Author's Affiliation Corporate Research Laboratory, Fuji Xerox Co., Ltd()
2nd Author's Name Masahiro MAEDA
2nd Author's Affiliation Corporate Research Laboratory, Fuji Xerox Co., Ltd
3rd Author's Name Noriji KATO
3rd Author's Affiliation Corporate Research Laboratory, Fuji Xerox Co., Ltd
4th Author's Name Hirotsugu KASHIMURA
4th Author's Affiliation Corporate Research Laboratory, Fuji Xerox Co., Ltd
Date 2004/6/11
Paper # PRMU2004-43
Volume (vol) vol.104
Number (no) 125
Page pp.pp.-
#Pages 6
Date of Issue