Electroencenphalogram-Based Effective Features for Sustained Attention Assessment in Conversation

Sustained attention, or concentration, is the ability to focus on a task for a long time. Its evaluation includes subjective assessments using questionnaires and objective assessments such as the electroencephalogram (EEG) method. This EEG method offers promising concentration level insights, but it is a challenge to fully leverage the multi-channel nature of EEG at the same time. In this study, we conducted a button-pressing task called a go/no-go task, designed to measure concentration to simulate a conversational scene while addressing the EEG multi-channel complexity. We measured response times and simultaneously recorded EEG. The extracted features such as frequency band characteristics and information-theoretical features like entropy were used to evaluate the concentration level. We employed group Lasso and cross-correlation analysis to expose the explainable EEG feature for sustained attention relevant to the VTC. As a result, we found that EEG power ratios were strongly related to concentration. By combining several power ratios, it may be possible to quantify the concentration.