Autoregressive modelling of electroencephalographic signals for medical simulators

Abstract

The recording of brain electrical activity has led to a greater understanding

of different neurophysiological states, has even made possible the

diagnosis of some neuronal disorders, hence the importance of characterization

and understanding of the different morphologies that may

have electroencephalography signals (EEG). The mathematical modeling

of biomedical signals facilitates the development of simulators that can

be useful as medical training tools on computers or mobile devices. This

paper presents the autoregressive (AR) modeling and simulation of

EEG signals in different physiological states: seizures, resting with eyes

open and eyes closed, and also under the presence of some of the most

common artifacts: muscle, eye blinking, electrode “pop”, and 60-Hz.

The performance of the models has been validated in the time domain

using the percentage of fitting (FIT), which was always above 70%, and

in the frequency domain through energy of the characteristic frequency

bands of the EEG. The modeling methodology, figures of simulated

signals and the values of the parameters evaluated are presented. The

wide variety of EEG signals modeled allow the development of brain

signals simulators for training of medical personnel, and also for the

analysis and characterization of EEG signals.