A method is proposed for the construction of empirical dynamical models for
processes with a high-dimension of the embedding space. It is shown that t
he higher dynamical degrees of freedom of a process observed can be taken i
nto consideration by generalizing third-order differential equations in the
class of equations with delayed argument. A method is proposed for the dec
omposition of the state vector of the observed process and for the effectiv
e estimation of the dimension of the embedding space. An example of the con
struction of the empirical model is presented for a quasi-stationary segmen
t of an actual electroencephalogram (EEG) of a human.