Administrating Cognitive Tests Through HRI: An Application of an Automatic Scoring System Through Visual Analysis

In this paper, we propose the use of a social robot to support professional figures in performing cognitive screening and stimulation. We implemented tools that allow us to automatize the evaluation of a subject’s cognitive abilities. More specifically, we programmed the humanoid robot Pepper to administer, in a fully automated and “friendly” way, three cognitive assessment tasks: Word List Recall (WLR), Attentive Matrices (AM), and the Rey-Osterrieth Complex Figure (ROCF). For WLR, we displayed the word list on the robot’s tablet, and the speech recognition module to record the recalled words. AM was delivered by asking the subjects to use and mark numbers on the tablet. For ROCF, we implemented two novel score assessment algorithms based on the processing of the picture drawn by the subjects and acquired through the robot camera. In particular, for ROCF, correlation analysis was conducted to compare automatically computed scores with a human psychologist’s. Our results suggest that the human psychologist’s workload can be reliably reduced thanks to the support of the robot.