Feature selection for attention demanding task induced EEG detection

Electroencephalography (EEG) is a popular noninvasive method used to record and analyse the electrical activity of the brain. Despite the poor spatial resolution, this tool provides a very high temporal resolution. With the use of a large number of channels, it is necessary to select the relevant features in EEG analysis. Hence, this research paper aims to identify the features that are capable of differentiating an attention-demanding task- induced brain activity from the resting state condition. Eleven different features including mean, root mean square, band power, skewness, mode, data range, interquartile range (IQR) and three Hjorth parameters are extracted from alpha, beta and gamma frequency bands of EEG. Each feature is tested using the statistical tool called paired t-test. Results demonstrate the importance of feature selection step for the recognition process. Hjorth parameters have shown significant statistical difference (p<; 0.05) between the datasets of attention task and resting-state and thus, can be a biomarker in this particular case.