Spectral EEG Biomarkers in Cognitive Characterization: Applications in VR, UXA, and Ergonomics

This study seeks to assess the applicability of EEG spectral biomarkers in application fields where cognitive characterization is required, e.g. Virtual Reality, User Experience Assessment (UXA), and Ergonomics. It aims to gauge users' cognitive states across varying task settings. We have gathered EEG data from three distinct datasets for this purpose. The first dataset encompasses EEG recordings from 36 participants under two conditions: at rest and while performing arithmetic operations. Additionally, participants were categorized as skilled or unskilled performers, making this dataset valuable for evaluating the effectiveness of different EEG features related to working memory. The second dataset comprises EEG data from 14 participants memorizing different quantities of characters (specifically, 2, 4, 6, or 8 characters) for three seconds. This dataset aims to replicate and assess how the identified biomarkers can distinguish between various levels of working memory within the same participant. The third dataset involves EEG recordings from 27 subjects engaged in a 90-minute Virtual Reality (VR) driving task, wherein they needed to maintain the car within the lane amid random deviations. This dataset serves the purpose of evaluating the descriptors' capacity to differentiate between states of high and low attention, as measured by their values before lane deviations. It also facilitates an exploration of how fatigue and time-on-task impact these markers. Our findings indicate that the Theta-to-Alpha ratio (TAR) measured at midline electrodes or as the ratio of frontal theta to parietal alpha effectively characterizes cognitive effort during mental arithmetic and memory tasks. In contrast, the Theta-Alpha-to-Beta Ratio (TA2BR) measured at temporal scalp locations emerges as the most efficient descriptor for assessing heightened vigilance states, particularly in tasks requiring external attention and rapid responses, such as the VR driving task. The influence of time-on-task on descriptor reliability varied depending on participants' performance levels.