Presentation | 2012-09-02 A Compensation Method of Motion Features with Regression for Defective Depth Image Ryu YUMIBA, Yoshiteru AGATA, Hironobu FUJIYOSHI, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | In this paper, we propose a method for compensating motion features utilizing regression estimate based oncorrelation between motion features from deficient human bodies and from entire ones, when recognizing actions from personswhose bodies are partially deficient. This compensation is good for the situation when human bodies are partially protrudingaround the edge of view angle, and contributes to enlarge region coverage of action recognition. Firstly in the proposed method,motion features and position are calculated from an acting person in a depth image. Secondly, deficit length is calculatedprotruding out of the view angle, according to the position of the person. Lastly, motion features from an entire body areestimated with regression estimate from the motion features above, selecting regression coefficients according to the deficitlength. Effectiveness for improving F-measure is confirmed with three kinds of motion features in a fundamental experiment ina laboratory. In the experimental results when human bodies are deficit from floor level to 600mm above the floor, F-measureis improved by more than 11.1% with the motion feature compensation comparing the case without compensation. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | Depth Image / Action Recognition / Regression Estimation |
Paper # | PRMU2012-35,IBISML2012-18 |
Date of Issue |
Conference Information | |
Committee | PRMU |
---|---|
Conference Date | 2012/8/26(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) | A Compensation Method of Motion Features with Regression for Defective Depth Image |
Sub Title (in English) | |
Keyword(1) | Depth Image |
Keyword(2) | Action Recognition |
Keyword(3) | Regression Estimation |
1st Author's Name | Ryu YUMIBA |
1st Author's Affiliation | Hitachi Ltd. Hitachi Research Laboratory Omika() |
2nd Author's Name | Yoshiteru AGATA |
2nd Author's Affiliation | Department of Computer Science Chubu University |
3rd Author's Name | Hironobu FUJIYOSHI |
3rd Author's Affiliation | Department of Computer Science Chubu University |
Date | 2012-09-02 |
Paper # | PRMU2012-35,IBISML2012-18 |
Volume (vol) | vol.112 |
Number (no) | 197 |
Page | pp.pp.- |
#Pages | 6 |
Date of Issue |