Simulation Of Spatiotemporal Pattern Embedded EEG And Its Analysis

Self-organized activities from synaptic interaction among populations of neurons form the sustained fluctuations of EEG. Based on this hypothesis, EEG is simulated via filtering the output of a random number generator.Based on the generation of correlated and uncorrelated noises, this paper simulates the aperiodic oscillatory waveforms of EEG, 1/f~α temporal power spectral densityand spatial power spectral density, nearly Gaussian amplitude histograms, and the distributions of eigenvalues under Principal Components Analysis. We impose 4 patterns of amplitude modulation in the simulated background EEG and classify them according to the Euclidean distance in 64-D space. We also convert the eigenvectors onto 2-D display space through Sammon's nonlinear mapping and classify one-to-one patterns among the 4 patterns using the Linear Discriminant Analysis. The KIII network and BP network are applied to recognize the patterns. The comparisons of these methods are made.Yet the study is primary, the simulation method presented in this paper shows the potential to provide test data to exam EEG analysis algorithms for neuropsychiatric disorder diagnosis and treatment evaluation, and the brain-computer interface studies to optimize the technique for extracting information about brain activity from EEG.