Application of machine learning in attentiveness detection from EEG signal

Abstract

Brain-computer interface (BCI) spellers enable severely motor-impaired people to communicate through brain activity without the use of their muscles. Our brains precisely predict what we will think. If a human-readable character can be identified by its appearance, our issues may be resolved. Currently, human, machine, and brain communication based on machine learning is highly believable. In this study, we intend to employ the non-invasive brain stimulation technique, often known as EEG, for the treatment of these individuals. A Braincomputer interface system based on electroencephalography provides the optimal solution to this issue. It establishes a link between the brain and the computer system, allowing brain waves to control our actions. The objective is to determine if a person is paying attention by recognizing characters from a dataset of P300, which is an event-related potential (ERP) component, using a BCI design. If a character is identified as a person paying attention, the data is labelled as target class; otherwise, the data is displayed as non-target. Our study has resulted in a number of Machine Learning strategy techniques. In this study, we analyzed the performance of four different types of Machine Learning Algorithms, including Logistic Regression (LRR), Random Forest Classifier, AdaBoost classifier, and XGBoost Classifier, to determine the most accurate algorithm. Custom CNN achieved the highest accuracy among classifiers, at approximately 88.46%.