Design Of Real-Time 3D Display Software For EEG Signals Based On ThinkGear ASIC Chip

Electroencephalogram(EEG)contains various brain activities such as movement,language,and consciousness which reflect changes in emotion and vigilance of human.With the development of hardware and software in the EEG monitoring,the software processing platform which satisfies command of researchers used to precisely extract and analyze EEG has developed rapidly.However,the evolution of real-time display of EEG signal processing seems slightly inadequate.In order to improve the practical value of the EEG display system,it is significant to improve the real-time performance of the display.In this thesis,a software platform which based on Think Gear ASIC combined with an efficient algorithm to implement emotion recognition of EEG signals for real-time three-dimensional(3D)display is designed.This platform consists of the acquisition system and the display platform.About the display platform,the Q3 DSurface class which renders 3D surface maps with multiple drawing modes and shadows in Qt was used in EEG signal surface maps so that the EEG data is transformed into easy-to-understand graphic and can be viewed by freely rotating.The main work of this article is as follows:(1)A real-time display platform for EEG signals is designed.The 3D power spectrum of EEG signals was displayed after Fast Fourier Transformation(FFT).The characteristics of brain waves can be obtained from the frequency from each fulcrum which contains visible frequency.(2)The EEG signal display platform combined with efficient algorithms to recognize emotion from different EEG signals.The thesis divides the algorithm into three steps.The first step is the noise removal process which uses filter processing to remove the noise signal from the time and frequency domains and the double-tree complex wavelet packet algorithm to decompose and reconstruct the EEG signal.The second step is feature extraction which is aimed at finding the sample entropy of the reconstructed signal and constructing sample characteristics as input to the support vector machine classification model.The third step is to train a classifier which establishes a support vector machine classifier to classify the sample characteristics.(3)Optimize the classifier.Aiming at the lower accuracy of sentiment classification accounting for the lack of a fixed method for selecting the parameters of the support vector machine,a hybrid algorithm,VGS-QPSO,is used to optimize the parameters of the support vector machine based on a variable mesh search method and a quantum particle swarm optimization algorithm.This algorithm greatly improves the efficiency and accuracy of support vector machine parameter optimization.