Cross-Sessions and Cross-Paradigms Analysis for the Problem of Brain-Computer Interface Inefficiency


Brain-Computer Interfaces (BCIs) allow users to make use of brain activity to control external devices directly for rehabilitation and enhancement of human functions. However, the inefficiency problem that a typical BCI system is unable to effectively decode EEG signals in some users, prevents BCI technology from benefitting all users. The proportion of inefficiency varies in the major BCI paradigms, among which Motor Imagery (MI)-based BCI achieves highest (10%-50%). Hence, the question arises as to whether other BCI paradigms, such as P300, could be substitutes for users who cannot be served by certain paradigm. In this work, a cross-paradigms BCI experiment, in which 93 healthy subjects executed BCI tasks including real movement and P300 for two sessions on separated days, was performed to answer the above question. Firstly, the highly correlation between the recognition accuracy in two sessions within subjects for both Sensory Motor Rhythm (SMR) features (p = 4.47×10 -11 ) and P300 features (p = 2.17×10 -3 ) indicated the reproducibility of the subject-level BCI inefficiency in the two paradigms. Further analysis demonstrated no significant correlation between the decoding performance of the SMR and P300 features (p = 0.604). The results verified the feasibility of improving BCI decoding performance by replacing certain BCI paradigm with another one when users encounter the problem of BCI inefficiency.

2021 27th International Conference on Mechatronics and Machine Vision in Practice (M2VIP)
Gan Huang
Gan Huang

My research interests include Neural Modulation, Brain Computer Interface and Neural Prosthetics.