System Design Of A Hybrid BCI Based On Eye-Tracking And SSVEP

In the last few decades,many studies have shown that humans can use brain signals to communicate with computers or machines by using Brain-Computer Interface(BCI).The BCI system can amplify and record the spontaneous,rhythmic electrical activities of brain cells through electrodes on the scalp and turn them into control signals for external devices.BCI is widely used in psychology,clinical diagnosis,cognition research.The hybrid BCI system is a multi-modal BCI system that integrating two or more different BCI systems to eliminate the disadvantages and retain the advantages of the single-modal one.The emergence of hybrid BCI has brought revolutionary significance to the field of BCI research.This study combined the Steady-State Visual Evoked Potential(SSVEP)from Electroencephalogram(EEG)and the eye tracking technology to build a novel hybrid BCI system.In the proposed hybrid BCI,the eye tracking technology was used for initial selection to obtain a subset of targets and SSVEP signals were used to identify the final target in the pre-selected subset.Compared with the traditional SSVEP-based single-modal BCI,the hybrid one obtained higher identification accuracy,information transfer rate(ITR)and provide better user experience.Compared with eye tracking technology,the hybrid one is able to avoid the "Midas Touch" problem and has higher precision.We firstly build a hybrid speller using the proposed hybrid BCI in this study.According to the results of online copy-spelling test,the hybrid speller achieved an average identification accuracy of 94.7%and an average ITR of 191.3 bits/min,but the SSVEP-based single-modal one is only 47.3%and 62.1 bits/min under the same conditions.We further increased the size of command set up to 112 and reduced the area of the buttons by 55.6%to test the horizontal scalability of the hybrid BCI.The results showed that the identification accuracy of the hybrid speller did not impacted significantly.Furthermore,we attempted to apply this hybrid BCI to a much more complex scenario,so a hybrid BCI web browser was implemented and an online test was performed to simulate daily Internet use.The results show that the proposed hybrid BCI web browser is not only user-friendly but also has a significant shorter target identification time and a significant higher ITR compared to other three BCI web browsers.Especially,the average target identification time is 8.09 seconds,which further reduced the gap between BCI and traditional mouse-keyboard interface.At present,the main goal of BCI research is to help improve the quality of life of patients with severe motor impairment through external devices.Generally speaking we not only need high accuracy and high ITR,but also require the reliability,usability and scalability of the system.The hybrid brain-computer interface proposed in this study has proved to be a zero-training,high speed,high scalable and user-friendly hybrid BCI,which can be used to help patients in need and bring them more confidence for life.