Digital Storage and Simulation of EEG-Data Using a Linear EEG-Model
A digitized EEG can be reduced to a few parameters (< 15) if an autoregressive model is introduced. These parameters can be stored very economically.With the aid of autoregressive parameters a recursive digital filter is realized which generates the EEG if a random number sequence is given to the filter input. Furthermore, the autoregressive parameters can be used for estimating power density distribution and for building up a non-recursive digital filter based on the inverse autoregressive model which will filter the EEG and detect spikes and other instationarities in the EEG. After decomposition of the recursive filter into a series of filters of second order it is possible to calculate parameters such as peak frequency, bandwidth and amplitude, which describe the second order filter and thus characterize the frequency components of the EEG.