Effective feature extraction is very important for motor imagery (MI) Electroencephalography (EEG) signal pattern recognition in brain machine interface (BMI) or brain computer interface (BCI) applications. Common spatial patterns (CSP) method is a frequently used machine learning algorithm for discriminative feature extraction in MI related BMIs. Although CSP is a famous and effective method in BMI applications, the intrinsic variations in the EEG signal properties would affect its performance. To address this issue, instead of using the eigenvalues for spatial filter vectors selection as instructed by the traditional CSP approach, we present a novel criterion which considers not only the differences between different classes but also the dissimilarities in the same class in a data-driven manner for robust spatial filter construction. We compared our method against CSP using a public dataset, namely dataset IVa of BCI competition III. The final results of the experiment indicate that our approach gives higher classification accuracy, by reducing the variations in the same class.
IOS Press, Inc.
6751 Tepper Drive
Clifton, VA 20124
Tel.: +1 703 830 6300
Fax: +1 703 830 2300 email@example.com
(Corporate matters and books only) IOS Press c/o Accucoms US, Inc.
For North America Sales and Customer Service
West Point Commons
Lansdale PA 19446
Tel.: +1 866 855 8967
Fax: +1 215 660 5042 firstname.lastname@example.org