A novel EMD-based Common Spatial Pattern for motor imagery brain-computer interface

Common Spatial Pattern (CSP) algorithm is a commonly used effective feature extraction method in motor imagery (MI) electroencephalogram (EEG) based brain computer interface (BCI). The motor imagery patterns extracted by CSP are associated with variations in subject-specific frequency bands power. Therefore, optimizing frequency band carrying MI intention is required by the CSP method. However, the frequency band is usually divided manually and evaluated which does make use of the EEG signal property and also with low efficiency. In this paper, we propose a novel Empirical Mode Decomposition (EMD) based CSP method to realize the data-related and adaptive frequency band selection. Specifically, the intrinsic mode functions (IMFs) decomposed from the EMD and the amplitude modulated signal by instantaneous amplitude (IA) calculated from Hilbert Transform have been fully explored and employed. Moreover, the intensive experiments have been conducted to evaluate the proposed method. From the experiment results, we observed that the EMD based CSP method enhances the classification accuracy in BCI competition IV dataset I for all subjects and a paired t-test shows a significant difference level.