The Identification Of EEG Signals And Its Application Research In BCI

Brain-computer-interface is a new communicate control system that converts EEG recorded from scalp into output control signals or commands, and it does not depend on the normal peripheral nerve and the muscle output channel. The input signal of the system is EEG, which is considered as some external characterization of various consciousness activity inside brain, and contains the rich information of cerebral cortex activities, and to some extent reflects brain activity. The processing and analysis of EEG signals play an important role on understanding and cognitive of brain function. So looking for the right EEG signal processing method and recognition algorithm is the main research contents in this paper, and it is also a key part of improving reliability and performance of brain-computer interface system.This paper mainly analyzed and discussed preprocessing and feature extraction and classification algorithm of EEG signals based on motor imagery, and the main contents are followed:(1) EEG theory is studied detailedly. EEG frequency band closely related with motor imagery is studied after analysing frequency information of EEG signals, and band-pass filter is designed to extract signals in the frequency range.(2) In this paper, noise removal methods of EEG signals based on independent component analysis (ICA) are studied. From ICA theory, the FastICA principle is studied, then FastICA is applied in multi-channel EEG signals, at last, EEG signals are decomposed into independent components to obtain the solution matrix. Noise can be identified through space model of the solution matrix. Research shows that ICA is able to eliminate the noise, enhance the useful information of the EEG signals and improve the SNR of EEG signals. ICA has good effectiveness in EEG signals denoising.(3) OVR-CSP is studied for multitasking identification problem and it is the extension of CSP. In this paper, we contribute by combing ICA with OVR-CSP. Finally, SVM is used as classifier model and EEG recognition rate is used for indexes. EEG signals are analysed by the denoising and feature extraction method (combine ICA with OVR-CSP). Results show that the proposed algorithm can improve the classification recognition rate of EEG signals.