Classification of EEG signals by multi-scale filtering and PCA

High accuracy for the classification of electroencephalogram (EEG) signal is an important basis for a brain-computer interface (BCI) system. In this paper, we proposed a novel approach to enhance the classification performance in identifying EEG signals, which classify EEG by combining multi-scale filters and principal component analysis (PCA). First, a multi-scale filter with different size of filter window was used to extract major frequency-band components from EEG signals. This might not only enhance the adaptability of filter to the EEG signals, but also satisfy the diversity of frequency resolution. Then PCA was utilized for feature extraction to reduce data dimension and improve the classification accuracy. The experimental results on EEG signals of motor imagery indicate that the proposed method is able to achieve a classification accuracy of 91.13%. Using this method might enhance the performance of a BCI system in signal recognition.