Human emotions detection based on a smart-thermal system of thermographic images

• Published in: Infrared physics & technology Vol. 81; pp. 250 - 261
• Main Authors: Cruz-Albarran, Irving A., Benitez-Rangel, Juan P., Osornio-Rios, Roque A., Morales-Hernandez, Luis A.
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.03.2017
• Subjects: Diagnostic; Emotion; Infrared thermography; Region of Interest; Self-calibration and biomarker; Infrared thermography; Region of Interest; Diagnostic; Emotion; Self-calibration and biomarker
• ISSN: 1350-4495; 1879-0275
• DOI: 10.1016/j.infrared.2017.01.002

•A study on thermal behavior of anger, disgust, fear, joy and sadness is carried out.•A self-calibrated system to have the same thermal trend for each subject is proposed.•A diagnostic of emotion through a top-down hierarchical classifier is done.•Biomarkers are proposed through temperature changes. This work presents a noninvasive methodology to obtain biomedical thermal imaging which provide relevant information that may assist in the diagnosis of emotions. Biomedical thermal images of the facial expressions of 44 subjects were captured experiencing joy, disgust, anger, fear and sadness. The analysis of these thermograms was carried out through its thermal value not with its intensity value. Regions of interest were obtained through image processing techniques that allow to differentiate between the subject and the background, having only the subject, the centers of each region of interest were obtained in order to get the same region of the face for each subject. Through the thermal analysis a biomarker for each region of interest was obtained, these biomarkers can diagnose when an emotion takes place. Because each subject tends to react differently to the same stimuli, a self-calibration phase is proposed, its function is to have the same thermal trend for each subject in order to make a decision so that the five emotions can be correctly diagnosed through a top-down hierarchical classifier. As a final result, a smart-thermal system that diagnose emotions was obtained and it was tested on twenty-five subjects (625 thermograms). The results of this test were 89.9% successful.