Frequency-Wavenumber Analysis of Eeg using Spatio-Temporal Autoregressive Model

Abstract

A new method has been proposed to estimate the frequency-wavenumber spectrum of sensory array data. This method is essentially an extension of the recently established method of estimating power spectrum by autoregressive analysis. The frequency-wavenumber spectrum analysis has been extensively studied in the field of seismology. Its applications in the biomedical field are still relatively few in number, although several authors have applied it to evoked responses in EEG to find out components which propagate with certain velocities and directions. The method proposed in this paper assumes a linear model for the generation of spatio-temporal EEG signals. Instead of directly computing the spectrum by FFT (Fast Fourier Transform) along with some window function, the parameters of the linear (autoregressive) model are estimated which are then used to obtain the spectrum. In this paper the new linear model (named spatio-temporal autoregressive model) is proposed and its relationship to the frequency-wavenumber spectrum is shown. Furthermore, some results of the application of this method to human EEG are presented.