Multiclass Classification of Brain-Computer Interface Motor Imagery System: A Systematic Literature Review

Abstract

The Brain-Computer Interface (BCI) is a great concept that enables people to interact with external devices solely through their brain signals. Motor imagery (MI), in which the acquired signals are captured from limb movements' imagination, is one of the most popular BCI research topics. For people with disabilities, this concept could be beneficial. The most common research for BCI MI classification has so far focused on a binary classification problem. In a real-world situation, however, the machine will need to train and differentiate more than two classes or solve a multiclass classification problem. Therefore, to summarize the research on multiclass BCI MI classification, this paper will conduct a systematic literature review for 30 articles that have gone through the selection process. This review found that the most used dataset in Multiclass BCI MI-EEG System is BCI Competition IV dataset 2a. As for the feature extraction method and classification method, most researchers used computationally inexpensive and stable methods. However, some of the researchers use more complex methods such as Fourier Transform as a feature extraction method and a Deep Learning-based classifier as a classification method.