Automatic estimation of cognitive load during robot-assisted gait training

Robot-assisted gait rehabilitation is becoming more and more common in patients with neurological impairments. Active patient participation in cognitively challenging t raining sessions are considered essential for the success of gai t rehabilitation [2], although objective assessments of cog nitive load are difficult to obtain during training. Questionnaires can be used to assess cognitive load, but only at discrete time points, often after training has ceased. Psychophysiological measurements have previously been used to infer to the psychological state of subjects, as every chang e in the psychological state has a physiological response [1] . Our objective was to determine if measurable physiological responses could be used to provide a continuous estimate of a subject’s cognitive load during robot-assisted gait training. We provided subjects with a virtual task during robot-assisted gait therapy to induce different levels of cognitive load. Physiological measurements in combination with machine learning techniques were used to objectively quantify the current cognitive load of the subject pe rforming the task in the virtual environment.