For the last portion of my disseration, I am aiming to model child affect through thermal imaging.
Child affect can be very difficult to determine as children can’t self-report their emotions. Employing biological signals such as the regulation of body temperature can help provide estimates of child affect during interactions with a robot.
We are employing a FLIR Lepton 3.5 with OpenCV and custom detection algorithms to locate landmarks of interest and correlate temperature values with affect through machine learning.
Associated Papers:
Ameer Helmi, Lara Rupnawar, Samuel W. Logan, and Naomi T. Fitter. Illuminating Engagement: Real-Time Thermal Imaging of Child Affect During Child-Robot Interactions. International Consortium on Rehabilitation Robotics (ICORR), Late Breaking Abstract Poster Presentation, September 2023.