Riemann Kernel Support Vector Machine Recursive Feature Elimination in the Field of Compound Limb Motor Imagery BCI

doi:10.3901/JME.2019.11.131

Abstract

Abstract: Compound limb motor imagery brain-computer interface (CLMI-BCI) has better rehabilitative potential after stroke than traditional motor imagery brain-computer interface (MI-BCI), because of its high complexity of instructions. However, it's ability of using for clinical is limited due to the low recognition accuracy. To solve this problem, a new method named Riemann kernel support vector machine recursive feature elimination (RKSVM-RFE) is proposed based on the manifold information on electroencephalogram (EEG). The EEG data of 10 subjects are collected when they were imagining 7-class movements of different parts of the body. The data is modeled using RKSVM-RFE to recognize the motor intention corresponding to the EEG data. Results show that accuracy from our method is about 7% higher than the state-of-the-art method named CSP. And RKSVM-RFE can reduce complexity of system because it can decrease 50% EEG channels. The research provides a new idea about the development of rehabilitation technology based on MI-BCI, which is worthy of further development.

Key words: BCI, feature selection, manifold, motor imagery, SVM

CLC Number:

R318

TAO Xuewen, YI Weibo, CHEN Long, HE Feng, QI Hongzhi. Riemann Kernel Support Vector Machine Recursive Feature Elimination in the Field of Compound Limb Motor Imagery BCI[J]. Journal of Mechanical Engineering, 2019, 55(11): 131-137.

References

[1] 陈树耿,贾杰. 脑机接口在脑卒中手功能康复中的应用进展[J]. 中国康复理论与实践,2017,23(1):23-26. CHEN Shugeng, JIA Jie. Application of brain-computer interface in rehabilitation of hand function after stroke (review)[J]. Chinese Journal of Rehabilitation Theory and Practice,2017,23(1):23-26.
[2] DALY J J,WOLPAW J R. Brain-computer interfaces in neurological rehabilitation[J]. Lancet Neurol,2008,7(11):1032-1043.
[3] KIM S P,SIMERAL J D,HOCHBERG L R,et al. Neural control of computer cursor velocity by decoding motor cortical spiking activity in humans with tetraplegia[J]. J. Neural Eng.,2008,5(4):455-476.
[4] BIRBAUMER N,WEBER C,NEUPER C,et al. Physiological regulation of thinking:Brain-computer interface (BCI) research[J]. Prog. Brain Res.,2006,159:369-391.
[5] JOHNSON S H. Imagining the impossible:Intact motor representations in hemiplegics[J]. Neuroreport,2000,11(4):729-732.
[6] JOHNSON S H,SPREHN G,SAYKIN A J. Intact motor imagery in chronic upper limb hemiplegics:Evidence for activity-independent action representations[J]. J. Cogn. Neurosci.,2002,14(6):841-852.
[7] ANG K K,CHUA K S,PHUA K S,et al. A randomized controlled trial of EEG-based motor imagery brain-computer interface robotic rehabilitation for stroke[J]. Clin. EEG Neurosci.,2015,46(4):310-320.
[8] 奕伟波. 复合运动想象诱发下的脑电响应机制与解码技术研究[D]. 天津:天津大学,2017. YI Weibo. Research on response mechanism and decoding technology of EEG induced by compound motor imagery[D]. Tianjin:Tianjin University,2017.
[9] CHATTERJEE R,BANDYOPADHYAY T. EEG based motor imagery classification using SVM and MLP[C]//International Conference on Computational Intelligence and Networks. 23-25 Dec. 2016. THDC Institute of Hydropower Engg and Technology,Bhagirathipuram,Tehri,India. IEEE,2016:84-89.
[10] LI D,ZHANG H,KHAN M S,et al. Recognition of motor imagery tasks for BCI using CSP and chaotic PSO twin SVM[J]. The Journal of China Universities of Posts and Telecommunications,2017,24(3):83-90.
[11] ARVANEH M,GUAN C,ANG K K,et al. Optimizing the channel selection and classification accuracy in EEG-based BCI[J]. IEEE Trans. Biomed. Eng.,2011,58(6):1865-1873.
[12] ALOTAIBY T,EL-SAMIE F E A,ALSHEBEILI S A,et al. A review of channel selection algorithms for EEG signal processing[J]. Eurasip Journal on Advances in Signal Processing,2015(1):1-21.
[13] ALEXANDRE B,STÉPHANE B,MARCO C,et al. Multiclass brain-computer interface classification by riemannian geometry[J]. IEEE Transactions on Biomedical Engineering,2012,59(4):920-928.
[14] BARACHANT A,BONNET S,CONGEDO M,et al. Classification of covariance matrices using a Riemannian-based kernel for BCI applications[J]. Neurocomputing,2013,112(10):172-178.
[15] CONGEDO M,BARACHANT A,KOOPAEI E K. Fixed point algorithms for estimating power means of positive definite matrices[J]. IEEE Transactions on Signal Processing,2017,65(9):2211-2220.
[16] RAMOSER H,MULLER-GERKING J,PFURTSCHELLER G. Optimal spatial filtering of single trial EEG during imagined hand movement[J]. IEEE Trans. Rehabil. Eng.,2000,8(4):441-446.
[17] MA Y,DING X,SHE Q,et al. Classification of motor imagery EEG signals with support vector machines and particle swarm optimization[J]. Computational and Mathematical Methods in Medicine,2016,1(10):1-8.
[18] GUYON I,WESTON J,BARNHILL S,et al. Gene selection for cancer classification using support vector machines[J]. Machine Learning,2002,46(1-3):389-422.
[19] LAL T N,SCHRODER M,HINTERBERGER T,et al. Support vector channel selection in BCI[J]. IEEE Trans. Biomed. Eng.,2004,51(6):1003-1010.
[20] GIBLIN P. A panoramic view of riemannian geometry by marcel berger[J]. Mathematical Intelligencer,2006,28(2):73-74.