Motor Imagery based Multivariate EEG Signal Classification for Brain Controlled Interface Applications

Brain computer interface (BCI) can be defined as a pathway that enables human brain to communicate and voluntarily command an external device and generate output instead of depending upon peripheral nerves and muscular movements. Achieving maximum classification accuracy is the greatest challenge in developing a BCI system to correctly interpret the brain signals. This paper aims at investigating various classification algorithms in combination with different pre-processing techniques and comparing their results for maximum classification accuracy. Independent component analysis (ICA), principal component analysis (PCA) and notch filters are used for artifact removal, dimension reduction and noise cancellation, respectively. Left and right hand movements were recorded from the scalp using non-invasive electrodes. Fine KNN, with independent components as feature, gives highest classification accuracy in comparison with various classification techniques used in this research.