This paper explores the process of self-guided learning of realistic facial expression production by a robotic head with 31 degrees of freedom. Facial motor parameters were learned using feedback from real-time facial expression recognition from video. The experiments show that the mapping of servos to expressions was learned in under one-hour of training time. We discuss how our work may help illuminate the computational study of how infants learn to make facial expressions.
10 aActuators 10 aEmotion recognition 10 aface detection 10 aface recognition 10 afacial motor parameters 10 aFeedback 10 aHumans 10 alearning (artificial intelligence) 10 aMachine Learning 10 aMagnetic heads 10 aPediatrics 10 areal-time facial expression recognition 10 aRobot sensing systems 10 arobotic head 10 aRobots 10 aself-guided learning 10 aServomechanisms 10 aServomotors 1 aWu, T. 1 aButko, N. 1 aRuvulo, P. 1 aBartlett, M. 1 aMovellan, J. uhttps://rubi.ucsd.edu/content/learning-make-facial-expressions 02225nas a2200373 4500008004100000020002200041245006700063210006300130260002900193520104300222653001901265653003001284653003801314653002801352653001901380653002101399653002201420653003801442653001101480653001901491653002001510653001701530653001301547653001901560653003001579653002001609653001501629653003601644653001901680653002001699100001801719700002501737856008901762 2008 eng d a978-1-4244-2153-4 00 aA discriminative approach to frame-by-frame head pose tracking 0 adiscriminative approach to framebyframe head pose tracking aAmsterdam bIEEE c09/2008 3 aWe present a discriminative approach to frame-by-frame head pose tracking that is robust to a wide range of illuminations and facial appearances and that is inherently immune to accuracy drift. Most previous research on head pose tracking has been validated on test datasets spanning only a small (< 20) subjects under controlled illumination conditions on continuous video sequences. In contrast, the system presented in this paper was both trained and tested on a much larger database, GENKI, spanning tens of thousands of different subjects, illuminations, and geographical locations from images on the Web. Our pose estimator achieves accuracy of 5.82deg, 5.65deg, and 2.96deg root-mean-square (RMS) error for yaw, pitch, and roll, respectively. A set of 4000 images from this dataset, labeled for pose, was collected and released for use by the research community.
10 aaccuracy drift 10 acontinuous video sequence 10 acontrolled illumination condition 10 adiscriminative approach 10 aface detection 10 aface recognition 10 afacial appearance 10 aframe-by-frame head pose tracking 10 aHumans 10 aImage analysis 10 aImage databases 10 aLaboratories 10 aLighting 10 aMagnetic heads 10 amean square error methods 10 apose estimation 10 aRobustness 10 aroot-mean-square error tracking 10 aSystem testing 10 aVideo sequences 1 aWhitehill, J. 1 aMovellan, Javier, R. uhttps://rubi.ucsd.edu/content/discriminative-approach-frame-frame-head-pose-tracking