Emotion Recognition from Non-Straight Walking Gaits Induced by Emotional Videos

Automatic emotion recognition is very useful in several situations. Human gait is an expression that can be observed at distance without subjects' awareness. This study shows the results of performing automatic emotion prediction with our previously proposed emotional gait dataset captured by OptiTrack motion tracking system. Microsoft HoloLens 2 smart glasses were used for showing emotional videos to induce subjects' emotions during walking. This study reveals that human gaits can be used for emotion prediction although the subjects walk in a non-straight walking path. Thresholding method was applied with leave-one-out cross validation. Emotions of walking subjects can be distinguished between happy and sad with an accuracy of 67.06%.