Automated Detection of Cognitive Load from Peripheral Physiological Signals based on Hjorth’s Parameters

摘要

The prolonged exposure to high levels of cognitive effort causes fatigue and stress-related decrease of attention and concentration, which are known to compromise work efficiency, safety, and health. In the present study, we investigate the applicability of the Hjorth parameters, namely Activity, Mobility, and Complexity, computed from peripheral physiological signals, as features on the automated cognitive load detection task. Specifically, here we consider the detection of high cognitive load in a person-independent scenario based on galvanic skin response (GSR) and photoplethysmographic (PPG) signals. To assess the practical worth of Hjorth’s parameters, we carried out a comparative evaluation in a common experimental protocol based on a subset of the CLAS dataset, which contains GSR and PPG recordings of 60 people while they were engaged in problem-solving tasks, such as Math-task and IQ-task. The discriminative capability of the Hjorth parameters was evaluated with four classification methods when Activity, Mobility, and Complexity are used individually and in combination. We report detection accuracy of up to 84.7% and 80.5% on the Math-task and IQ-task, respectively.