Unobtrusive estimation of psychological states through human movement observation

The purpose of this study was to investigate the possibility of estimating human emotions and three other psychological states based on external signals such as human body movements. To protect subjectsʹ privacy and minimize intrusiveness, the use of video was limited and no devices were attached to the subjects. Instead, motion and infrared sensors were utilized to estimate emotion states and other human condition in real-life settings. The results of the experiment demonstrated that walking velocity and motion load descriptors were good predictors of neutral or sad states. Future studies should improve the emotion estimation methodologies to prove effective in a larger group of subjects and to include a larger set of emotions.