A comparison of two digital methods of EEG simulation

Abstract

This article applies two very different approaches to the problem of digital simulation of a stationary time function, and compares their results. The first technique approxi mates the sampled time function by a Markov process of small order. The transition probabilities are estimated from samples of actual EEG, and a waveform is generated from these probabilities, using random number techniques.

The second technique treats a time function as the out put produced by a linear filter on some random input. The transfer function of the filter and the distribution of the in put signal are then calculated.

Graphs of the autocorrelation functions of actual EEG and both types of simulation are included for comparison. In the author's opinion, these graphs show the filtered- noise simulation to be somewhat superior to the Markov process simulation.