Stress and anxiety detection using facial cues from videos. (January 2017)

• Record Type: Journal Article
• Authors: Giannakakis, G.; Pediaditis, M.; Manousos, D.; Kazantzaki, E.; Chiarugi, F.; Simos, P.G.; Marias, K.; Tsiknakis, M.
• Is Part Of: Biomedical signal processing and control. Volume 31(2017)
• Page Start: 89
• Page End: 101
• Publication Date: 2017-01
• Subjects: Facial cues -- Emotion recognition -- Anxiety -- Stress -- Blink rate -- Head motion; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2016.06.020
• Languages: English
• ISSNs: 1746-8094

Abstract

Graphical abstract: Highlights: A framework for stress detection through video recorded facial cues is proposed. A thorough experiment using external and internal stressors is established. Analysis relies on semi-voluntary facial features instead of facial expressions. Relationship between participants self reports and facial signs is estimated. Facial cues implicated in stress/anxiety are identified using machine learning techniques. Abstract: This study develops a framework for the detection and analysis of stress/anxiety emotional states through video-recorded facial cues. A thorough experimental protocol was established to induce systematic variability in affective states (neutral, relaxed and stressed/anxious) through a variety of external and internal stressors. The analysis was focused mainly on non-voluntary and semi-voluntary facial cues in order to estimate the emotion representation more objectively. Features under investigation included eye-related events, mouth activity, head motion parameters and heart rate estimated through camera-based photoplethysmography. A feature selection procedure was employed to select the most robust features followed by classification schemes discriminating between stress/anxiety and neutral states with reference to a relaxed state in each experimental phase. In addition, a ranking transformation was proposed utilizing self reports in order to investigate the correlation of facial parameters with a participant perceived amount of stress/anxiety. The results indicated that, specific facial cues, derived from eye activity, mouth activity, head movements and camera based heart activity achieve good accuracy and are suitable as discriminative indicators of stress and anxiety.