A review on features extraction and classification methods of EEG-based brain-computer Interface

The Computer Brain Interface (BCI) is an effective communication system for people withneurological disorders who do not need external muscular activity. Extraction and classificationof main and key steps in BCI systems based on motor imagery. In this review article, FFT, WPD,CSP and GC feature extraction methods introduced and reviewed. The Common Spatial Pattern(CSP) is an efficient and common technique for extracting data properties used in BCI systems. Inaddition, LDA, SVM, NN and DL classification methods introduced and reviewed. LDA and SVM(linear) methods are the most common linear classification algorithms used in BCI systems. Forfeature vectors with high dimension, SVM is the most appropriate classifier due to its insensitivityto curse of dimensionality. In recent years, DL used in the design of BCI systems. DL is a goodchoice for BCI systems based on motor imagery with big dataset.

Keywords:

Brain computer interfaces (BCI) , Common spatial patterns (CSP) , Electroencephalography (EEG) , Feature selection , Classification