Behind-the-ear EEG for SSVEP-based BCIs

Abstract

Locked-in syndrome is a neurological disease which limits patients from executing any muscular activity. This in turn restricts such patients from expressing their needs and ideas to the rest of the world. Numerous approaches have been considered to use electroencephalographic (EEG) activity, in order to create a communication link between the brain and the external world, without engaging in any muscular activity. This has been achieved through brain-computer interface (BCI) systems. This study consists of the analysis of steady state visual evoked potentials (SSVEPs), which are EEG signals recorded while subjects are focusing on stimuli flickering at specific frequencies. Throughout this study, different datasets were used in order to determine whether SSVEPs can also be detected from the behind-the-ear area, as compared to the occipital area, and whether these are effective enough to design a more practical BCI system which does not necessary require wearing a headset with electrodes. While reviewing other research papers on signal processing techniques used in SSVEPbased BCI systems, pre-processing, spectral analysis, feature extraction and classification techniques were chosen accordingly. Canonical correlation analysis (CCA), which is one of the standard techniques used in this domain, was chosen for the extraction of features. Additionally, a literature review on studies involving the analysis of brain signals recorded from the occipital and behind-the-ear regions was carried out, on which the study of this project was based. Using open source datasets, the classification accuracies of both occipital and behind-the-ear regions were compared, concluding that the classification performance when using behind-the-ear regions is inferior to that obtained when using the occipital regions and this difference is on average greater than 20%. Analysis was also carried out per subject for both occipital and behind-the-ear regions. It was concluded that for most of the subjects, the classification performance at the behind-the-ear region deteriorated by more than 40% from that obtained from the occipital electrodes. However, there were few of the subjects that obtained a deterioration in the classification performance between the occipital and behind-the-ear regions of around 25% only. Therefore, a behind-the-ear BCI system would be worth to use it on such subjects.