Application of Neural Network Technologies in Signal Detection Tasks

The paper considered the problem of applying neural network technologies to detect signals in conditions of complex non-Gaussian disturbances with a priori uncertainty of probability distributions. In particular, the synthesis and comparative analysis of two approaches to the detection of a harmonic signal with a known frequency and an unknown phase against the background of correlated autoregressive interference was carried out. The first approach is based on the procedure of statistical synthesis of the adaptive signal detection algorithm. The second approach involves the use of a neural network in combination with the Fourier transform. It was noted that neural network algorithms significantly outperform optimal algorithms in modeling the action of non-Gaussian impulse disturbances, but demonstrate lower efficiency compared to the action of Gaussian correlated interferences.