Feature Selection for EEG-Based Fatigue Analysis Using Pearson Correlation

Mental fatigue is one kind of exhaustion that occurs in a person's mental state. Mental fatigue will arise if the brain is continuously forced to work. This mental fatigue is also common to happen to the senior high school student, especially in Indonesia because mostly they attend the school as a full day school. This study is exploring the mental fatigue condition in 13 senior high school students who attend a full day school by using EEG by selecting the appropriate feature for recognizing the mental fatigue. Recently, EEG technology has been implemented by some studies in the past to explore mental fatigue. In this study, EEG recording is held in the morning and carried out without stimulation. Meanwhile second measurement is used as a test condition. In the second test, the EEG recording was held in the afternoon, and stimulation of the arithmetic test was given to induce the fatigue. Baseline conditions describe the condition of fresh, while the second test conditions describe the condition of fatigue. The feature extraction process was done in time domain by analyzing 4 features: mean, mean absolute, standard deviation, and the number of zero crossing. Pearson correlation was applied to select the features by ranking the correlation between baseline conditions and test conditions (R-value=1). Using those data, F-Test calculation is done on each band (Alpha, Beta, Tetha) to verify fatigue condition. Based on P-Value from F-Test calculation, we conclude Beta and Theta band showed a significant increase during fatigue condition (P-Value < 0.05) compared to alpha band.