In this paper we propose new approaches to the control of mathematical models of electroencephalogram signals, and also discuss ways to adjust their parameters. The developed models are based on the simulation of postsynaptic impulses using a combination of Gaussians with random amplitudes and mean values. The width parameter is estimated by calculating the convolution of the electroencephalogram signal with the Gaussian and subsequent statistical analysis of the results. The adequacy of the resulting model was verified using neural networks, random forest, and method of expert assessments.