Temporal Features and Machine learning Approaches to Study Brain Activity with EEG and ECG

Electroencephalogram (EEG) and Electrocardiogram (ECG) are electrical signals that reflect activities of the brain and cardiac, respectively, based on which some neurological disorders and mental status are determined. In this paper, a novel set of temporal features like energy, Shannon energy, entropy, and temporal energy, all together along with machine learning-based classifiers to identify the relaxing state of humans and while performing mental tasks like arithmetic operations using these signals are proposed. Machine learning approaches, such as k-nearest neighbors, support vector machine, decision tree, gradient booster, logistic regression, and random forest are utilized for classification between these two states. A publicly available dataset at physionet.org that includes 36 subjects with almost half of adult males and females are used in this study. On this dataset, which includes 21 channels (20 EEG + 1 ECG), the Random Forest algorithm performed best compared to others with an accuracy of about 99.34%. The proposed set of features was capable to yield a slightly better result than state-of-the-art.