Thinking Eeg And P300 Eeg Feature Extraction And Recognition

Brain Computer Interface(BCI)is a real time communication system that connects the human brain with the external devices.As the transmitted signal in the BCI system,Electroencephalogram (EEG)is always mixed with many kinds of artifacts,such as EOG,ECG,EMG,baseline,etc.The background noise,which is consisted ofαrhythm andβrhythm and so on,is part of artifacts to the basically weak P300 signal.It poses a challenge in EEG signal process that how to extract useful information from the raw EEG signals and to find the efficient pattern and classify algorithm.Recently,it's proved that the Independent Component Analysis(ICA)is an efficient algorithm applied in Blind Source separation(BSS).Wavelet Transform is a useful filter in application.More application with the two algorithms is taken in biomedical signal process field.The thesis investigates removing noise from mental tasks signal and P300 EEG with the two algorithms,and discussing how to classify the movement imagery EEG.Following is the works we have finished.1.Discussion on remove noise from mental tasks EEG with ICA is posed.In order to remove the baseline artifact,two more sine signals with the same frequency of the baseline are plugged in.By setting certain independent component zeros,it can cut off the noise.Then separate the needed signals from the EOG artifact as above.2.Introduce a method to classify movement imagery signals by calculating the signals' second order moment(energy).When human beings actually or image to move their left or right hand,its neural activity in the primary sensor motor areas of the same side will rise and the opposite one will reduce.That's the special characteristic of mu EEG.To calculate the second order moment is easy and fit for the BCI system on line analysis.The analysis of raw EEG dynamic property verities this pattern can reach satisfied result.When we picks 448 samples(lasts 3.5s),the result can achieve 86.43%.3.Detection of event related potentials P300 signal de-noising with ICA and Wavelet multi-resolution decomposition.ICA can separate the hidden independent components from the multiple channel source signals.Based on this,we can remove the physiology artifacts like EOG, EMG and background artifacts such asαrhythm andβrhythm.Then we apply Wavelet multi-resolution decomposition to reserve the needed band of P300 EEG.It's verified to be efficient in extracting P300 signals.Finally,research on how to discriminate the P300 EEG is discussed.