Högskolan i Skövde

Remote photoplethysmography (rPPG) can be used to remotely estimate heart rate (HR) of users to infer their emotional state. However natural body movement and facial actions of users significantly impact such techniques, so their reliability within contexts involving natural behavior must be checked. We present an experiment focused on the accuracy evaluation of an established rPPG technique in a gaming context. The technique was applied to estimate the HR of subjects behaving naturally in gaming sessions whose games were carefully designed to be casual-themed, similar to off-the-shelf games and have a difficulty level that linearly progresses from a boring to a stressful state. Estimations presented mean error of 2.99 bpm and Pearson correlationr = 0.43, p < 0.001, however with significant variations among subjects. Our experiment is the first to measure the accuracy of an rPPG techniqueusing boredom/stress-inducing casual games with subjects behaving naturally.