出版社：Academy & Industry Research Collaboration Center (AIRCC)
摘要：When automatic recognition of emotion became feasible, novel challenges has evolved. One ofthem is the recognition whether a presented emotion is genuine or not. In this work, a fullyautomated system for differentiation between spontaneous and posed smiles is presented. Thissolution exploits information derived from landmark points, which track the movement offiducial elements of face. Additionally, the smile intensity computed with SNiP (Smile-Neutralintensity Predictor) system is exploited to deliver additional descriptive data. The performedexperiments revealed that when an image sequence describes all phases of smile, the landmarkpoints based approach achieves almost 80% accuracy, but when only onset is exploited,additional support from visual cues is necessary to obtain comparable outcomes.
