Two detection algorithms are compared when they use the energy detection criterion - the "classical" problem of testing two simple hypotheses and the application of a filter based on truncated order statistics (TOS-filter), which replaces the integrator in the "classical" problem. On statistical models, the influence of the parameters of the TOS-filter on its output characteristics was studied. It is shown that when using the Neyman-Pearson criterion, the cuto􀁸 threshold for the TOS-filter is determined by the median of the noise. The dependence of the detection probability on the number of integration time partitions in the energy receiver is studied. The dependence of the detection probability on the size of the sliding window matrix is studied. It is shown that the use of the TOS-filter provides a significant gain in the detection probability.