Application Of Independent Component Analysis In EEG Signal Processing

With the global brain research continuous warming,the various methods of brain science have gradually become a hot topic in the natural science domain.EEG collected from the human scalp as an intuitionistic and nondestructive reflection of the electrophysiological activity occurring in the course of brain information processing,for its unique attributes (non-invasive,high time resolution,etc.),is increasingly becoming an essential research tool of related research and provides important technical support for in-depth development of brain science.In the EEG signals,theory is generally believed that interference(power frequency,ECG artifact,eye movement artifact,etc.) and evoked potentials arising from a specific stimulation are signals from different sources,can be considered mutual statistical independence in terms of time.Signals from different sources in the brain tissue are almost instantaneous linear mixture,superposition delay effect and convolution effect can be ignored.Various artifacts and EEG signals are usually subject to non-Gaussian distribution.At the same time,we assume that the number of EEG signals observation channels is equal to the number of sources.For the defects that traditional artifact subtraction method will inevitably remove some components of EEG signals and superposed average method does not reflect the successive changes of evoked potentials,this dissertation uses independent component analysis to solve the problem of artifacts removal and feature extraction in EEG signals from a new perspective. From the independent components of the decomposition,we combinate its spatial distribution pattern to identify which components reflect artifacts or evoked potentials.In the artifacts removal,we will track the independent components which reflect artifacts to zero and return other components to the original EEG signals by multipling the inverse matrix of mixing matrix at the left to achieve the purpose of artifacts removal.In the extraction of evoked potentials,we only return the independent components which reflect evoked potentials to the original EEG signals by multipling the inverse matrix of mixing matrix at the left to achieve the purpose of extraction.Finally,we study the effectiveness of this method used in the artifacts removal and extraction of evoked potentials through the simulation test of real EEG signals.The results show that this method can effectively remove a wide variety of EEG artifacts,and there is no damage to other components of EEG signals.Two types of evoked potentials extracted at one trial in line with the physiological significance of evoked response.