Fatigue EEG Feature Extraction Based on Tasks With Different Physiological States for Ubiquitous Edge Computing

Abstract

Mental fatigue is a gradual and cumulative phenomenon that manifests in the weakening of human physiological activities for ubiquitous edge computing in the Internet of Things. In this paper, two groups of Stroop tasks with different difficulty levels are proposed to induce fatigue, which is evaluated via electroencephalogram (EEG). Wavelet packet decomposition and sample entropy algorithm are utilized to analyze the EEG signals in both sober and fatigue state. The experiment results show that compared with the sober state, the fatigue state has a higher $\alpha $ wavelet relative energy and $\theta $ wavelet relative energy and significantly lower $\beta $ wave relative energy ( $P ). The ratio of parameters $\alpha /\beta $ and $(\alpha +\theta)/\beta $ increases with the fatigue degree, and the sample entropy of each brain region shows a decreasing trend. Compared with the more difficult task group, the change of parameters in the low-difficulty task are more obvious. Hence, the suggested parameters $\alpha /\beta $ and $(\alpha +\theta)/\beta $ can be used as potential indicators to measure mental fatigue, and the appropriate increase in the difficulty of the tasks may be inversely related to the generation of mental fatigue to some extent.