DefinePK

Abstract

The Brain Computer Interface (BCI) systems have incredible applications both in clinical and non-clinical areas. Electroencephalography (EEG) is one of the most used neuroimaging techniques to acquire brain activity in BCI Systems. However, EEG signals are usually very complex and requires extensive processing to analyze them. This paper explores the implementation of motor imagery (MI) paradigm based BCI system based on the on deep learning. A typical deep learning model includes the stages of pre-processing, feature extraction and classification in single model. However, such model requires lot of data for training purpose. In order to compensate this data requirement, this paper implements a deep learning model based on CNN with extracted features as an input. The implemented model consists of three CNN layers followed by fully connected layers. The model performed with 80% of classification accuracy on average in offline analysis. In real-time analysis, the approximate accuracy was 66.9 % across the subjects.