A Novel Movement Intention Detection Method for Neurorehabilitation Brain-Computer Interface System

• Published in: 2018 IEEE International Conference on Systems, Man, and Cybernetics (SMC) pp. 1016 - 1021
• Main Authors: Song, Minsu, Oh, Senghue, Jeong, Hojun, Kim, Jongbum, Kim, Jonghyun
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01-10-2018
• Subjects: Brain-computer interface; Data mining; Electroencephalography; False positive; Motor imagery; Robots; Selectivity; Sensitivity; Task analysis; Timing; Training data
• ISSN: 2577-1655
• DOI: 10.1109/SMC.2018.00181

In brain-computer interface based on motor imagery for rehabilitation, false positive could be a major cause of undesired brain plasticity which ends up with the wrong reconstruction of damaged brain tracts. Moreover, the number of electroencephalogram (EEG) electrodes required would be the reason of practical difficulties to clinical use. To reduce the false positive and the number of electrodes required, we proposed a novel two-phase classifier based on detecting Mu band event-related desynchronization (ERD). Along with five channels to detect motor imagery, the algorithm only uses three channels to reject ERD-like noise or non-motor signals. The performance of the proposed algorithm was evaluated through two-day experiments with four healthy subjects. The total sensitivity was 60.83% and the total selectivity was 78.49%. Those experimental results show that the proposed method can reduce the rate of false positives with small number of EEG channels.