Authors:
Yang Fu
and
Claude Frasson
Affiliation:
University of Montreal, Canada
Keyword(s):
Emotion Recognition, IAPS, Skin Temperature, Thermal Emotional Profile, Machine Learning, EEG, HMM (Hidden Markov Model), Infrared Camera.
Related
Ontology
Subjects/Areas/Topics:
Affective Computing
;
Biomedical Devices for Computer Interaction
;
Biomedical Engineering
;
Computer Graphics and Visualization of Physiological Data
;
Devices
;
Health Information Systems
;
Human-Computer Interaction
;
Methodologies and Methods
;
Physiological Computing Systems
;
Video and Image Analysis for Physiological Computing
Abstract:
Human can react emotionally to specific situations provoking some physiological changes that can be detected using a variety of devices, facial expression, electrodermal activity, and EEG systems are among the efficient devices which can assess the emotional reactions. However, emotions can trigger some small changes in blood flow with an impact on skin temperature. In the present research we use EEG and a thermal camera to determine the emotional profile of a user submitted to a set of emotional pictures. Six experiments were performed to study the thermal reactions to emotions, and in each experiment, 80 selected standard stimuli pictures of 20 various emotional profiles from IAPS (a database of emotional images) were displayed to participants every three seconds. An infrared camera and EEG were used to capture both thermal pictures of participants and their electrical brain activities. We used several area of the face to train a classifier for emotion recognition using Machine Lea
rning models. Results indicate that some specific areas are more significant than others to show a change in temperature. These changes are also slower than with the EEG signal. Two methods were used to train the HMM, one is training classifier per the participant self data (participant-independent), another is training classifier based on all participants` thermal data (participant-dependent). The result showed the later method brings more accuracy emotion recognition.
(More)