Imaging the Finger Force Direction Yu Sun, John Hollerbach and Steve Mascaro

1. x y z Imaging the Finger Force Direction Yu Sun, John Hollerbach and Steve Mascaro ysun@cs.utah.edu, jmh@cs.utah.edu, smascaro@mech.utah.edu Background • Force --> color • Color --> force • Color pattern -> force direction • People have the same bone, tissue structure LDA extracted features Elastically deform fingernail to atlas Nail to nail and skin to skin LDA extracted features – nail only Image Alignment Source images 1. Intrafinger alignment • Reference – new image • Feature detection Corners, Edges • Feature correlation • Correct pairs • Wrong pairs-outlier • RANSAC • Random sample • Remove outlier by fitting model • 2D homography mapping Training data projected on the first two feature vectors Elastically deformed images Feature extraction and classifier 2. Interfinger alignment Classification results on new images References: Y. Sun, J.M. Hollerbach, and S.A. Mascaro, "Imaging the fingertip force direction," IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR), June 18-23, 2007, Minneapolis, MN, in press. Y. Sun, J.M. Hollerbach, and S.A. Mascaro, "Finger force direction recognition by principal component analysis of fingernail coloration pattern," World Haptics Conference, Tsukuba, Japan, March 22-24, 2007, in press. Y. Sun, J.M. Hollerbach, and S.A. Mascaro, "EigenNail for finger force direction recognition," Proc. IEEE Intl. Conf. Robotics and Automation (ICRA), April 10-14, 2007, Rome, Italy, in press. 6 groups for forces: +x, -x, +y, -y, z, zero, Comfortable force, force level was not specified 420 training images 840 verification images LDA feature extraction with training images • Segment the nails: Canny edge filter • Cubic B-spline close boundary • Atlas: regular shapes