Development of Single-Channel Hybrid BCI System Using Motor Imagery and SSVEP.

Li-Wei Ko,S S K Ranga,Chung-Chiang Chen,Oleksii Komarov

doi:10.1155/2017/3789386

Abstract

Numerous EEG-based brain-computer interface (BCI) systems that are being developed focus on novel feature extraction algorithms, classification methods and combining existing approaches to create hybrid BCIs. Several recent studies demonstrated various advantages of hybrid BCI systems in terms of an improved accuracy or number of commands available for the user. But still, BCI systems are far from realization for daily use. Having high performance with less number of channels is one of the challenging issues that persists, especially with hybrid BCI systems, where multiple channels are necessary to record information from two or more EEG signal components. Therefore, this work proposes a single-channel (C3 or C4) hybrid BCI system that combines motor imagery (MI) and steady-state visually evoked potential (SSVEP) approaches. This study demonstrates that besides MI features, SSVEP features can also be captured from C3 or C4 channel. The results show that due to rich feature information (MI and SSVEP) at these channels, the proposed hybrid BCI system outperforms both MI- and SSVEP-based systems having an average classification accuracy of 85.6 ± 7.7% in a two-class task.

Highlights

A brain-computer interface (BCI) establishes a human-todevice communication channel by translating the brain signals into machine codes to control external devices or applications [1, 2]
A hybrid BCI system is a combination of a primary BCI system with another communication channel, which can be a BCI or another system based on a physiological signal recognition like electromyography (EMG) and electrooculography (EOG)
We focus on technological progression of a hybrid BCI system using motor imagery (MI) and state visually evoked potential (SSVEP)

Summary

Introduction

A brain-computer interface (BCI) establishes a human-todevice communication channel by translating the brain signals into machine codes to control external devices or applications [1, 2]. Numerous techniques for feature extraction methods [4,5,6], classification algorithms [7, 8], and experimental paradigms [9, 10] have been developed. The majority of these systems were based on a single modality of EEG, that is, they either use motor imagery (MI) [9], P300 [10], or steady-state visually evoked potential (SSVEP) [11]. We focus on technological progression of a hybrid BCI system using MI and SSVEP

Objectives

Methods

Results

Conclusion