Support Vector Machines And Its Application On Eeg Inverse Problem

Brain electricity phenomenon is one of external manifestations which the cranial nerve activitys.To extract the useful information from EEG will enable people to read rapid, non-invasive and economic vivo brain function (physiology, pathology, psychology, perception, cognition, etc.). The computer, the information processing method and the computation electromagnetics'application, brings the new vigor for the brain electricity research. The core technology on EEG research field is the study of the forward, inverse problem. The EEG inverse problem got extensive attention in the past 20 years. An exact result of this issue has important scientific significance and clinical application on cognitive function in brain research.Support vector machine algorithm based on the structural risk minimization (SRM) principle, rather than the traditional statistical experience risk minimization (ERM), specifically for the limited sample study, the situation could have been avoided local maximum, and through the non-linear transformation ingenious solved the dimension of the problem. Many have demonstrated superior performance methods in accuracy over traditional learning algorithm or with the equally. Support vector machine algorithm in pattern recognition, regression estimation, the estimated probability density function, and other aspects of the application and made a lot of applied research results to promote the development in various fields.In this work, we will use support vector machine algorithm to solve the dipole source localization of inverse problem. From the study of the forward problem'model and learning algorithm we get the fundamental understanding of EEG problem. From the study of the inverse problem'characteristics we analyze the existing algorithms'strengths and shortcomings in solving the inverse problem'. And then, we introduced SVM classification algorithm and regression algorithm and its various optimization algorithms, and analysis of multi-SVM classification algorithm. We realize that the SVR learning algorithm based on statistical samples to find the law and the law on the use of statistics to predict unknown data, the algorithm EEG inverse problem for the dipole source location without any prior knowledge, without in the process of being solved, the reliability of positioning with unknown dipole will not increase. It is a good direction to present SVR in solving inverse problem'source localization.