Brain-Computer Interface System Based On Gaussian Mixture Model Classification Research Of EEG Signals

In recent years, along with the world gradually accelerating pace of population aging, the number of people with disabilities continue to rise, forcing the human brain should accelerate the pace of scientific research, people aspire to the study of brain science and related fields to overcome the shortcomings and deficiencies of human beings. People are eager to direct with a large thinking of brain activity signals to communicate with the outside world, People eager to use brain signals to communicate directly with the outside world, even control the surrounding environment. Brain-Computer Interface(BCI) provides a scientific way to achieve this dream. Human hope this new communication technology can be used to assist traffic control tools, weapons and other systems, particularly for those with neuromuscular impairment, in patients who are unable to use the normal means of communication to provide a way to communicate with the outside world, so that people with disabilities and limb movements difficult to regain the ability of older people to communicate freely with the outside world. BCI is a new way of human-computer interface, is a hot topic in recent years, the study of brain function.For the brain-computer interface system, there are two main forms of international, one is the study of a class of non-invasive brain-machine interface, and the other is an in-vasive study of brain-computer interface as a result of non-invasive brain-machine interface. Because of non-invasive brain-machine interface subjects did not hurt, so broad range of ap-plications, without invasive brain-machine interface based on EEG brain-machine interface and because of its simplicity and safety award Domestic and foreign researcher’s attention. EEG classification problem is brain-machine interface system, a process critical to the ac-curacy of the classification directly affects the brain-machine interface system performance. The most commonly used method is Bayesian classification(Bayes) linear classifier, which specializes in two-class problems, but when dealing with multi-class classification of EEG data is not very satisfactory, algorithm execution is slow. To avoid this problem, on the basis of previous research, we propose the use of Gaussian mixture model(GMM), it can effective-ly deal with multi-class data clustering problem, when a multi-class EEG data clustering, mixed Gaussian model with a Gaussian model assumes that each category corresponding. The paper first choose principal component analysis(PCA) method for feature extraction of EEG data, the extracted principal components as a Gaussian mixture model clustering features of the object; then use expectation-maximization(EM) algorithm for Gaussian mix-ture model parameters to estimate; final accuracy of the model in terms of clustering, we introduce the field of operations research is a very important algorithms:Hungarian algo-rithm, so that the data can be found on the EEG The optimal matching of the Gaussian model (cluster). Finally, the classification results are given Gaussian mixture model and Bayes and Gaussian mixture model based on the average of the accuracy of clustering over 87.288%. while the Bayes average of linear classifier accuracy is 74.501%. The results show that Gaussian mixture model with the Hungarian algorithm effectively improve the accuracy of EEG clustering and suitable for multi-class clustering EEG.