Multi-class classification for intuitive MI-BCI in 3D spaces

Abstract

Brain-computer interface (BCI) has been drawing increasing attention from the researchers in recent years. Motor imagery (MI) based BCI is one of the most studied categories of BCIs, which extract features from cortical activities during the imagination of various body movements. With the unique advantages of self-motivated and asynchronous protocol in MI-BCI, by translating brain activities into meaningful control commands over external devices, MI-BCIs are making the concept of mind control a reality. The performance of recent MI-BCI systems are impressive, however, the widely used MI commands are lack of mental connection to the actual movements of the external systems. To solve this issue, a new MI-BCI pipeline that implements intuitive MI commands is proposed in this article, and the performance of the system is evaluated in a 6-class classification tasks in 3D spaces. After considering natural motions in real-life scenarios that are related to directional control, we selected the variation of driving in our study. The results showed a remarkable mitigation of mental workload using the intuitive MI tasks, and the classification accuracy of the proposed system was satisfying. The successful attempt of using intuitive movements for MI-BCI proved the feasibility of the customization of MI task selection for individual research.