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

• Published in: 2022 10th International Conference on Affective Computing and Intelligent Interaction Workshops and Demos (ACIIW) pp. 1 - 8
• Main Authors: Jianwattanapaisarn, Nitchan, Sumi, Kaoru, Utsumi, Akira
• Format: Conference Proceeding
• Language: English; Japanese
• Published: IEEE 18.10.2022
• Subjects: Emotion Induction; Emotion recognition; Gait Analysis; Legged locomotion; Motion pictures; Non-straight walking Gait; Shape; Smart Glasses; Tracking; Trajectory; Turning
• DOI: 10.1109/ACIIW57231.2022.10086022

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%.