A SPA-based Manifold Learning Framework for Motor Imagery EEG Data Classification
This work addresses the problem of accurate EEG classification for brain-computer interface systems, particularly with small sample sizes, but it is incremental as it builds on existing manifold learning and CSP methods.
The paper tackles the challenge of classifying motor imagery EEG data with limited subjects by proposing a manifold learning framework using a spherical approximation classifier, achieving high accuracy on the 2008 BCI competition data and improving decoding accuracy with strong robustness for small datasets.
The electroencephalography (EEG) signal is a non-stationary, stochastic, and highly non-linear bioelectric signal for which achieving high classification accuracy is challenging, especially when the number of subjects is limited. As frequently used solution, classifiers based on multilayer neural networks has to be implemented without large training data sets and careful tuning. This paper proposes a manifold learning framework to classify two types of EEG data from motor imagery (MI) tasks by discovering lower dimensional geometric structures. For feature extraction, it is implemented by Common Spatial Pattern (CSP) from the preprocessed EEG signals. In the neighborhoods of the features for classification, the local approximation to the support of the data is obtained, and then the features are assigned to the classes with the closest support. A spherical approximation (SPA) classifier is created using spherelets for local approximation, and the extracted features are classified with this manifold-based method. The SPA classifier achieves high accuracy in the 2008 BCI competition data, and the analysis shows that this method can significantly improve the decoding accuracy of MI tasks and exhibit strong robustness for small sample datasets. It would be simple and efficient to tune the two-parameters classifier for the online brain-computer interface(BCI)system.