Learner Cognitive State Recognition Based on EEG Signal Feature and Channel Selection

With the development of personalized education, accurately detecting learners’ cognitive state is crucial to improve learning outcomes and develop higher-order abilities. Learners display varying cognitive states during different cognitive engagement activities, which are linked to changes in brain activity. Further research is required to determine how to select effective electroencephalogram (EEG) features for cognitive state classification, despite previous studies on the detection of cognitive states based on EEG signal features. This paper investigates feature selection using the sequence forward feature selection algorithm based on Random Forest (RF-SFS) and channel selection using the Genetic Algorithm (GA) to detect the effects of relevant features and channels on cognitive state classification. The aim is to improve the efficacy of cognitive state detection. First, we designed learning activities that trigger different cognitive states. Then, subjects were recruited to participate in the activities. Next, we extracted EEG features from 64 electrodes and used RF-SFS for feature selection. The selected features were then input into various machine-learning classifiers to select effective features and identify the most relevant waveforms. The results of the classification indicate that the mean, first-order difference mean, and first-order difference standard deviation features of gamma bands demonstrate excellent performance in detecting the cognitive state of learners, with an accuracy of 99.86%. Subsequently, GA based on Random Forest was employed to select the most relevant channels. The research demonstrates that by selecting only 15 EEG channels in the gamma band, an accuracy rate of 99.05% can be achieved in the classification of students’ cognitive state. The feature selection method effectively screens the most relevant EEG features, improving the effectiveness of cognitive state detection. These findings provide support for the design of EEG devices for cognitive state detection.