Signal-Based Cognitive State Analysis for Adaptive E-Learning Environments

Abstract

The adaptive e-learning systems are increasingly based on the behavioural interaction data to personalise the delivery of the content, but this data can only give limited information on the cognitive state of the learner taken. This paper is aimed at addressing this shortcoming by proposing a signal-based cognitive state analysis model where real-time adjustment of e-learning negotiable settings can be made with the application of physiological cues to detect alterations. It is based on a closed-loop adaptive learning system with the proposed methodology combining the signal acquisition, preprocessing and feature extraction with a data-driven model of cognitive state. Raw physiological data are initially filtered and converted to obtain discriminative characteristics of time and frequency with which to describe changes in learner attention and mental workload. Those characteristics are then employed to train a supervised learning model to infer discrete candidate mental states which in turn are translated into adaptive learning interventions based on rule constrained decision logic. Experimental analysis on a controlled learning dataset shows that the suggested signal-based model shows high cognitive state classification than baseline methods that use traditional interaction measures. Moreover, it was found that cognitive-based adaptation leads to a study improved engagement and learning efficiency of learners, which indicates the usefulness of incorporating the analysis of physiological signals into adaptive e-learning systems. The suggested framework offers a scale-up and expansion framework of intelligent learning environments facing the next-generation that need precise cognitive cognizance and dynamism in real-time.