Stacked probabilistic regularized LDA on partitioning non-stationary EEG data for left/right hand imagery classification

Electroencephalography (EEG) is a user interface for communicating with patients, especially for tasks such as classifying Left/Right hand to represent YES/NO. Although there were many proposed classification techniques, none of them considered the non-stationary characteristic of brainwaves; thus, they cannot really be employed in real-world situations. In this paper, we aim to tackle the non-stationary issue by capturing all change points in EEG signal and partitioning it into a set of signals. Then Regularized LDA is applied to each of the partitioning signals, resulting in stacked of classifiers. The experiment was conducted on the BCI Competition IV (2b) data set. Our proposed method was compared to nine baseline techniques in four groups: (i) static algorithms, (ii) variants of adaptive Pool Mean algorithms, (iii) Adaptive CSP, and (iv) Corrected Sequential EM. The results showed that our method yields higher accuracy and significantly outperformed all baselines.