Based On Wavelet And Independent Component Analysis Of Eeg Signal Processing

Electroencephalogram (EEG) or brain wave is a typical kind of bioelectricity signals. EEGs are brain electrical potentials recorded by electrodes placed on the scalp of subjects. So EEGs contain large amount of original information about the activity of the human brain. The studying of EEG signals can help us to know well the connections between the electrical activity of neuron and the physiological and/or psycho-physiological significance. The analysis of EEG is important both for clinical medicine and for cognitive science research.But, the EEG signal that we can acquired is very weak and is badly contaminated by strong background noise, such as electrooculogram (EOG), electrocardiogram (ECG), and line noise (50Hz or 60 Hz power frequency interference), etc. EEG is a typical non-stationary random signal with a certain extent of non-Gaussian and non-linear character. The traditional analysis method generally considers the signal as linear quasi-stationary Gaussian distribution. In some cases, the Gaussian assumption may cause the results unacceptable and impractical.Modern information processing approaches such as Wavelet Transform (WT) and Independent Component Analysis (1CA) associated with the characteristic of EEG are investigated in this thesis. ICA from the angle of multi-dimensional statistic to process the signals, WT can study the signal at different resolution scale. Many methods applying the ICA and the WT are discussed in detail for removing and/or reducing the noise or artifacts from the observed EEG. The primary experimental results show that these methods can work effectively, especially for multi-channel EEG, and also can mine some hidden features from the EEG.The innovated works we have finished are as follows:1. Present the methods using WT or !CA to remove artifacts in EEG signals under different circumstance;2. Study how to combine the WT and ICA to enhance or extract the typical feature in the EEG;3. Discuss the ICA Gradient Adaptive Algorithm and apply it to EEG de-nosing on-line;4. Make some comparisons between the methods discussed in this paper and the traditional method.