MI 2 MI: Training Dyad with Collaborative Brain-Computer Interface and Cooperative Motor Imagery Tasks for Better BCI Performance

TL;DR

This study introduces a collaborative brain-computer interface training method using cooperative motor imagery tasks with a humanoid robot, enhancing individual BCI performance through dyadic interaction.

Contribution

It presents a novel cooperative MI training paradigm with a humanoid robot, demonstrating improved EEG quality and MI classification accuracy for individuals.

Findings

Enhanced EEG signal quality after cooperative training

Improved MI classification accuracy post-training

Dyadic collaboration boosts individual BCI performance

Abstract

Collaborative brain-computer interface (cBCI) that conduct motor imagery (MI) among multiple users has the potential not only to improve overall BCI performance by integrating information from multiple users, but also to leverage individuals' performance in decision-making or control. However, existed research mostly focused on the brain signals changes through a single user, not noticing the possible interaction between users during the collaboration. In this work, we utilized cBCI and designed a cooperative four-classes MI task to train the dyad. A humanoid robot would stimulate the dyad to conduct both left/right hand and tongue/foot MI. Single user was asked to conduct single MI task before and after the cooperative MI task. The experiment results showed that our training could activate better performance (e.g., high quality of EEG /MI classification accuracy) for the single user than single MI task, and the single user also obtained better single MI performance after cooperative MI training.