KMID : 1144120140040020149
|
|
Biomedical Engineering Letters 2014 Volume.4 No. 2 p.149 ~ p.157
|
|
Classification of motor imagery tasks for electrocorticogram based brain-computer interface
|
|
Xu Fangzhou
Zhou Weidong Zhen Yilin Yuan Qi
|
|
Abstract
|
|
|
Purpose: In the present study, we propose a novel scheme for motor imagery (MI) classification of multichannel electrocorticogram (ECoG) recordings from patients with medically intractable focal epilepsy.
Methods: This scheme proposes a combination of the two features which includes autoregressive (AR) model coefficients and local binary pattern (LBP) operators. It can provide spatial resolution and angular space information. Then the gradient boosting (GB) in conjunction with ordinary least squares (OLS) algorithm is employed as the classifier to improve the performance of MI classification for ECoG based Brain Computer Interface (BCI) system.
Results: Experimental results on the BCI Competition III data set I indicate that the novel method has excellent performance and yields a cross-validation accuracy of 88.8% and accuracy of 93%, respectively.
Conclusions: From the experimental results and comparative studies, we can infer that the scheme may serve as a good MI classification tool for a better tradeoff between the classification accuracy and computational complexity.
|
|
KEYWORD
|
|
Local binary pattern (LBP), Brain-computer interface (BCI), Electrocorticogram (ECoG), Motor imagery (MI)
|
|
FullTexts / Linksout information
|
|
|
|
Listed journal information
|
|
|