Artificial Intelligence Methods for a Brain-Computer Interaction Based on Electroencephalograms

Abstract

Electroencephalogram (EEG) brain-computer interface systems, particularly those that employ Motor-Imagery (MI) signals, have shown promise in controlling electromechanical devices. A suitable candidate for BCI systems is EEG because of its ease of recording and lack of invasiveness. To let people interact with their environments, BCI systems that use MI calculate a neural activity and translate these electrical impulses into effects or gestures. This research provides a concise overview of methods utilised to analyse EEG signals within the last ten years. Acquisition, pre-processing, feature extraction, and classification are the four facets of EEG signals in BCI systems covered in detail in this study. When it comes to extracting features from EEG-BCI systems, the most used time-frequency approach is Wavelet Transform (WT), and its upgraded version is Wavelet Packet Transform (WPT). Research on the categorization of motor imagery signals for BCI systems using ML and DL approaches was prompted by the advancements in artificial intelligence technologies. The study's authors hope that by sharing their findings, researchers will be able to locate reliable Machine Learning and Deep algorithms for feature extraction, which would aid in the development of a reliable EEG-BCI system.