This article is devoted to time-frequency signals analysis algorithm.This algorithm introduce the approach based on behavior functions and arithmetic series. The basis of p-adic numbers will be used to describe the discrete signal values. It will allow to build system behavior functions as a distribution of possibility measure. The function data analysis allows to perform the meta systems identification and build impulse functions. These functions will be used for estimation of frequency spectrum of initial signal. The study results of the algorithm performance on non-stationary signals model are given.