Modern special-purpose radios utilize ultra-short burst mode with frequency hopping spread spectrum. Such digital radios operate, as a rule, in the fast frequency hopping spread spectrum mode and use a considerable number of frequency channels with frequency separation between adjacent channels, which is significantly less than the spectrum width of the frequency element. The main modulation in such radios is frequency manipulation. The identification of signals with fast frequency hopping spread spectrum and frequency manipulation is performed by estimating the speed of frequency tuning, the speed of information transmission, the number of frequency channels and the step of the frequency grid and comparing the obtained values with the corresponding characteristics of known radio stations with fast frequency hopping spread spectrum. The time limits and the duration of the frequency elements are calculated by the envelope of the received signal. For this purpose, a suitable method has been developed, which consists in filtering the envelope by means of a moving average window for suppression of the noise component and threshold processing. The threshold value is selected by the given probability of its exceeding by noise sample. The proposed approach to determining the time limits of frequency elements does not require any a priori signal information. To determine the frequencies of the frequency elements periodogram estimates are used. An exponential extrapolation of the discrete power spectrum was used to increase the frequency resolution at a fixed window length of the fast Fourier transform. It is shown that in order to uniquely determine the step of the frequency grid and the diversity of frequency manipulation frequencies, it is necessary to analyze the number of frequency elements not less than the number of frequency channels. By analyzing the histogram of the differences of sorted frequency vector of the frequency elements, determine the frequency grid, frequency diversity and the multiplicity of frequency manipulation. An approximate estimate of the number of frequency channels is calculated as the ratio of the magnitude of the frequency vector of the frequency elements to the step of the frequency grid. As a rule, symbol transmission at frequency manipulation is carried out at orthogonal frequencies, so it can be roughly assumed that the symbolic speed is equal to the frequency diversity of the frequency manipulation. The baud rate can be set only after signal demodulation and bit stream analysis, which can determine the frame structure and the number of service and information bits. The proposed method provides the estimation of the values of the parameters necessary for the identification of radio stations of signal parameters with a relative error not exceeding 0.3%, with values of signal to noise ratio not lower than 5 dB.