Signal processing techniques for motor imagery brain computer interface: A review

Motor Imagery Brain Computer Interface (MI-BCI) provides a non-muscular channel for communication to those who are suffering from neuronal disorders. The designing of an accurate and reliable MI-BCI system requires the extraction of informative and discriminative features. Common Spatial Pattern (CSP) has been potent and is widely used in BCI for extracting features in motor imagery tasks. The classifiers translate these features into device commands. Many classification algorithms have been devised, among those Support Vector Machine (SVM) andLinear Discriminate Analysis (LDA) have been widely used. In recent studies, the researchers are using deep neural networks for the classification of motor imagery tasks. This paper provides a comprehensive review of dominant feature extraction methods and classification algorithms in brain-computer interface for motor imagery tasks. Authors discuss existing challenges in the domain of motor imagery brain-computer interface and suggest possible research directions.