PCA와 LDA를 이용한 EEG 신호 분석에 관한 연구

A brain-related biological signals, such as EEG (Electroencephalogram), are known to be important factors for accurate BCI (Brain-Computer Interface) system research. Among others, EEG signals are widely used to research the brain activity, because of their non-invasive, convenient properties. In addition, There are used to people who have normal imagination, but do not handed as Lou Gehrig’s disease. For all the advantages, EEG signals have a critical noise-related limitation. In this paper, a novel method using PCA (Principal Component Analysis) and LDA (Linear Discriminant Analysis) is proposed to classify the EEG signals accurately. By using PCA and LDA, characteristic vectors of the extracted EEG signals are visualized and it can reduce the dimension of the input signals effectively. The result of reduction of dimension is useful for classification process of EEG signals. The classification experiment of motor imaginary prove the efficiency of the proposed method. BCI Competition IV database that provided EEG raw data of motor imaginary is used in this experiment.