Hemodynamic Analysis for Cognitive Load Assessment and Classification in Motor Learning Tasks Using Type-2 Fuzzy Sets
Lidia Ghosh, Amit Konar, Pratyusha Rakshit, Atulya K. Nagar · IEEE Transactions on Emerging Topics in Computational Intelligence · 2018
This paper addresses a novel approach to assess and classify the cognitive load of subjects from their hemodynamic response while engaged in motor learning tasks, such as vehicle driving. A set of complex motor-activity-learning stimuli for braking, steering-control, and acceleration is prepared to experimentally measure and classify the cognitive load of the car drivers in three distinct classes: high, medium, and low. New models of general and interval type-2 fuzzy classifiers are proposed to reduce the scope of uncertainty in cognitive load classification due to the fluctuation of the hemodynamic features within and across sessions. The proposed classifiers offer high classification accuracy over 96%, leaving behind the traditional type-1/type-2 fuzzy and other standard classifiers. Experiments undertaken also offer a deep biological insight concerning the shift of brain activations from the orbitofrontal to the ventrolateral prefrontal cortex during the high-to-low transition in cognitive load. Furthermore, the activation of the dorsolateral prefrontal cortex is also reduced during the low cognitive load of subjects. The proposed research work outcome may directly be utilized to identify driving learners with a low cognitive load for difficult motor learning tasks, such as taking a U-turn in a narrow space and motion control on the top of a bridge to avoid possible collision with the car ahead.