Classification of Cognitive Load from Task Performance & Multichannel Physiology during Affective Changes

Detecting cognitive load (CL) is important for implementing adaptive systems aware of the of such system is to reduce error related risks during task-critical operations or to facilitate computer based (e.g. tutoring systems) learning. Self reports, task performance, behavioral and physiological changes are some of the indicators of CL. However, affective factors which occur naturally in everyday life can influence these variables, making CL detection challenging. In this paper, we investigate machine learning techniques for detecting CL levels from task performance and physiological channels during affective changes with the support of systematic experimental setup. Classification results are presented for detecting two levels (low & high) of CL from individual modalities and their combination using a vote classifier. We have achieved around 66% to 78% accuracy for detecting the two CL levels under affective changes (valence/arousal).