Potential of Pupil Diameter Monitoring for the Detection of Affective Changes in Human-Computer Interaction

There is continued interest in mechanisms that could inform computers of the affective state of their users. Some proposed approaches analyze physiological signals from the users to asses their affective states. Most of these solutions focus on physiological signals traditionally used for “detection of deception” (lie detector) experiments, such as the Galvanic Skin Response (GSR), the Blood Volume Pulse (BVP), or the Skin Temperature (ST). We monitored these signals, as well as the Pupil Diameter (PD), during a sequence of congruent and incongruent Stroop trials, expected to result in intervals of relaxation and stress in the subject (respectively). Single feature signals were derived from these variables (GSR mean, BVP interbeat interval mean, ST slope and PD mean) and used as detection signals for the identification of incongruent Stroop segments (i.e., stressed states in the subject). Receiver Operating Characteristic (ROC) analyses were carried out for each of the four single detection signals, as a mechanism to compare their discriminating power. The PD signal was found to result in the higher area under the ROC curve, which may imply a stronger potential for the discrimination of the stressed segments elicited in the subject by incongruent Stroop stimulation.