Font Size: a A A

Studies On EEG-based Cognitive Dynamics System Using Linear/nonlinear Methods And Dynamical Time-frequency-space Analysis

Posted on:2005-03-05Degree:DoctorType:Dissertation
Country:ChinaCandidate:X B MiaoFull Text:PDF
GTID:1104360125963623Subject:Electrical engineering
Abstract/Summary:PDF Full Text Request
A new concept, cognitive dynamics system (CDS), is presented in this dissertation, which argues that studies on brain cognitive function should be based on dynamic and systematic schemes. Since there always exists a working loop when our brains fulfill cognitive tasks, namely, a working loops of "information input (via seeing, listening, feeling, etc.) -information archiving (via memory) - information processing (via calculation, judgment, decision, etc.) - information output (via speaking, writing, emotion, etc.)", furthermore, such working loops are not static but labile, evolutionary, nonstationary, nonlinear, and chaotic. The synchronization and integration of these working loops at different time points, deferent brain regions, and different frequencies forges the whole cognitive functionalities of the neural system. Thus hybrid schematic methods including linear/nonlinear methods and dynamical time-frequency-space analysis are needed to meet the spatial, temporal, and frequency variability of cognitive process.Two groups of experiments were designed for CDS, one is under normal condition, and the other is under pathological condition. EEGs of 8 cognitive states under normal condition (eyes closed, eyes open, mental arithmetic, digit span, contrary digit span, orientation test, vocabulary association, graphic reasoning) were recorded on 20 healthy volunteers aged between 18 and 20 years individually; EEGs of 2 cognitive states under pathological condition (eyes closed, eyes open) were recorded on 20 cerebral hemorrhage and infarct patients aged between 47 and 84 years individually. For each EEG recording 16 Ag-AgCl disc electrodes were placed on the frontal (Fpl, Fp2, F3, F4, F7, F8), central (C3, C4), parietal (P3, P4), temporal (T3, T4, T7, T8) and occipital (O1, O2) locations according to the international 10/20 system, and were referenced against linked earlobes. The cut-off frequencies of the EEG amplifiers were set to 0.3 and 120 Hz with sampling frequency of 1000 Hz (12 bit).In the spectral analysis, FIR band-pass filtering and weak stationarity test were employed for data preprocessing, then Burg AR model based spectral estimation was discussed in comparison with Welch methods in details. Classical spectral indices (spectral edge frequency, median frequency, peak power frequency, rate, relative power rate of etc.) were analyzed under different cognitive states.In the nonlinear dynamics analysis, a novel ICA-based spatial-temporal filtering was designed for data preprocessing companying with weak stationarity test. Time delay and embedding dimension, dynamics parameters for phase space reconstruction, were selected using mutual information method and Cao's method individually. Then, three classic dynamics indices, correlation dimension, largest Lyapunov exponent, and approximate entropy, and their algorithms were discussed and analyzed in details under different states.In the dynamical time-frequency-space analysis, this new method and its significant roles in the CDS were presented firstly. Then two basic dynamical time-frequency-space analysis based on Gabor Transform and Wavelet Transform and their analyzed results under different cognitive states were discussed in details. Finally, another new dynamical time - frequency -space analysis called wavelet-chaos method was tentatively proposed to meet the nonlinearand nonstationary nature of EEG signals.In the EEG synchronization analysis, EEG synchronization methods were introduced systematically, including linear synchronization (cross correlation and coherence), nonlinear dynamics based synchronization (nonlinear interdependencies, synchronization likelihood, and cross approximate entropy), and phase synchronization (Hilbert Transform based and Wavelet Transform based methods). Then cross correlation synchronization, nonlinear interdependencies, Hilbert Transform based phase synchronization for a and band, cross approximate entropy and their characteristics were analyzed and discussed in details under different cognitive states.An im...
Keywords/Search Tags:Cognitive dynamics system, spectral analysis, nonlinear dynamics analysis, dynamical time-frequency-space analysis, synchronization analysis
PDF Full Text Request
Related items