Optimal set of EEG electrodes for real-time cognitive workload monitoring

Real-time assessment of cognitive workload underlies various applications that aid users´ smarter life style in terms of mental health and learning. Previous studies have shown that cognitive workload can be estimated from EEG. In practical applications using cognitive workload, it is crucial to make the sensor device more essential while maintaining performance at acceptable level. However, little is known about the number and positions of EEG electrodes that can provide robust performance in diverse situations. To this end, we carried out an analysis using a widely used Emotiv EPOC to identify optimal sets of EEG electrodes robust to differences between individuals, tasks, and sessions. Seven healthy subjects participated to perform two widely used working memory tasks: n-back and mental arithmetic task. We prioritized EEG channels based on the summarized importance of individual features from each channel. By subsequently adding the highly ranked channels, we could achieve classification accuracy of 80% on average for both working memory tasks in an online analysis with 7 channels.