Deep feature normalization using rest state EEG signals for Brain-Computer Interface

The brain-computer interface (BCI) system provides information exchanges between neural signals containing the user's intention and device control signals. Electroencephalogram (EEG) is a widely used signal for obtaining neural signals. In EEG decoding, EEG variability across different subjects critically degrades deep learning performance. In this paper, we propose a feature normalization method for reducing EEG variability with rest state EEG signals. The decoding structure is trained with a normalized feature which is normalized by subtracting the normalization feature extracted from the normalization structure. Experimental results show that the deep feature normalization algorithm dramatically enhances the performance of conventional deep learning algorithms.