Feature extraction from EEG data for a P300 based brain-computer interface

Brain-computer interface (BCI) is an input method that helps users to control a computer system using their brain activity rather than a physical activity that is required when using a keyboard or mouse. BCI can be especially helpful for users with limb disabilities or limitations as it does not require any muscle movement and instead relies on user’s brain activity. These brain activities are recorded using electroencephalogram (EEG). Classification of the EEG data will help to map the relevant data to certain stimuli effect. The work in this paper is aiming to find a feature extraction technique that can lead to improve the classification accuracy of EEG based BCI systems that are specifically designed for incapacitated subjects. Through the experiments, the implementation of Independent Component Analysis (ICA) and Common Spatial Pattern (CSP) extracted features from P300 based BCI EEG data and it was found that ICA and CSP produce more discriminative feature sets as compared to raw EEG signals. © 2017, Springer International Publishing AG.