LumEDA: Image Luminance Based Contactless Correlates of Electrodermal Responses

This work builds upon the findings of a 2020 pilot study which for the first time showed that correlates of electrodermal activity (EDA) could be estimated using simple spatial filters on video data captured under special lighting conditions. A total of 25 participants were recruited for this study in which they were presented with visual and auditory stimuli, video sequences and sensor-based EDA recorded from the left hand. A contactless estimate of the EDA was calculated using only the grayscale information from the video. Data from 18 participants was analyzed and it was seen that the average responses from the camera-based contactless estimate and the ground truth EDA were strongly positively correlated for both visual (Pearson r = 0.84, p < 0.01) and auditory stimuli (Pearson r = 0.86, p < 0.01). Most previous studies involve stronger stimuli such as breathing exercises, Valsalva Maneuver or self-pinching. This study shows that camera-based methods are also able to capture responses to relatively weaker stimuli. However, large inter-subject variability was also observed across the participants. A camera-based method not only allows for contactless estimation of EDA, but also facilitates spatiotemporal analysis of the distribution of electrodermal activity over a larger area of the body.