Measuring Cognitive Load Through Event Camera Based Human-Pose Estimation

The cognitive load is related to the effort associated with performing a specific task. It can affect learning efficiency, problemsolving abilities, and overall performance. In human-robot collaboration, the capability of a robot to assess cognitive load of humans during a joint task execution can positively affect performance, as robots can adapt its behaviour and mitigate cognitive demands. In this study, we measured behavioural responses to cognitive load, to assess how varying levels of cognitive load affects human body pose and movement. To this aim, we used low-latency, high temporal resolution, compressive event cameras and a lightweight human pose estimation network that can work in realtime at high frequency. Our results demonstrate that participants exhibit stiffer body movements under high cognitive load and more relaxed movements when cognitive load is low. Also, the reaction time is affected by the level of cognitive load. These findings suggest that cognitive load can be effectively inferred from event-driven pose estimation data, offering a non-invasive real-time method to monitor cognitive states and implement online adaptive responses in collaborative tasks.