Bilateral adaptation and neurofeedback for brain computer interface system

Brain computer interface (BCI) provides an alternative communication pathway between human brain and external devices without the participation of peripheral nerves and muscles. Although the BCI techniques have been developing quickly in recent decades, there still exist a number of unsolved problems, such as instability, unreliability and low transmission rate in real time applications of BCI. In the present study, we design a bilateral training framework for both human and the BCI system to improve recognition accuracy and to reduce the impact caused by non-stationary EEG signal. The statistical analysis is used to test whether there is an obvious improvement in recognition performance after using the bilateral adaptation strategy. The statistical analysis indicates that our algorithm is significantly different from the existing method in both conditions of trials (p = 0.0073) and sliding time windows (p = 0.00077). The results of statistical analysis reconfirm that performance using our algorithm is distinctly improved. The online experiments also demonstrate that the proposed algorithm achieves higher prediction accuracy and reliability compared with the existing method. The objective of our research is to transfer this strategy to some practical applications (e.g., electrical wheelchair control) for the better performance.

Research highlights▶ The BTP's model is updated in real-time by the EEG data from the current and previous sessions. ▶ The 'pure' trial selection. ▶ Multiple models are trained and their results are combined to predict. ▶ Subjects are able to adjust immediately their own brain activities according to neuro-feedback.