An Agent Based Approach for Electroencephalographic Data Classification

Brain Computer Interfacing (BCI) has become an emerging trend in the domain of ascertaining alternative approaches to interact with computers. The significance of BCI is better realized in the context of physically disabled persons. A typical BCI system recognizes the intentions of a patient (user) employing an Electroencephalography (EEG) device attached to his/her scalp. The EEG technology basically measures the electrical activity of the human brain (existent as an electrical potential), available on specific positions of the scalp, using electrodes. Since the electrical activity of the human brain is an abstract representation of the regional brain activity, divergent ongoing brain functions can be detected with this impressive technology. These functionalities, corresponding for distinct intentions of the user, are mapped to different interaction commands to be executed in the BCI system. EEG data classification for recognizing distinct user intentions is therefore, fundamental to achieve the goal of implementing an accurate BCI system. We have preferred a consumer-grade EEG device Emotiv's `Insight', which follows a non-invasive approach, for EEG data acquisition. The major reason of selecting the particular device was to align with our primary objectives, implementing a cost-effective (but high-accurate) BCI system. The key challenge associated with low-cost consumer-grade devices is the limited accuracy. Nevertheless, we have obtained an imposing precision of 76.6 %, which is remarkable, when compared to the maximum of 60.2 % achieved in the class of EEG devices. We are confident about the cause behind this attainment: concerning Multi-Agent Systems (MAS), an Artificial Intelligence (AI) technique, for EEG data classification.