Proc. SPIE. 11055, XII Conference on Reconnaissance and Electronic Warfare Systems
KEYWORDS: Signal detection, Continuous wavelet transforms, Stochastic processes, Wavelets, Signal processing, Signal to noise ratio, Digital signal processing, Artificial neural networks, Defect detection
The report examines the issue of increasing the efficiency of detecting complex impulse stochastic signals in the process of their generation against the background of quasi-periodic deterministic interference by using wavelet transformations and neural networks. An example the detection of a triple-wave stochastic signal is considered. One of the most characteristic signs of the shape of such signals is a sharply expressed asymmetry: the amplitude of the negative part of the signal is usually 3-4 times higher than the positive maximum amplitude. The second very important feature is the ratio of the positive parts amplitudes of the signal: the amplitude of the right-hand side is always greater, or in extreme cases, equal to the amplitude of the left-hand side. The proposed technique for processing such impulse signals against a background of quasi-periodic interference by using wavelet-neural technologies for analyzing digital signals. For this purpose, an artificial neural network was constructed, which made it possible to detect such signals at the beginning of their development, starting from a signal-to-noise ratio of 1.5 times, which is twice as good as the threshold for visual analysis. The proposed technique can be used in the analysis of pulsed signals in radar systems, mobile railroad rail diagnostic systems by the Magnitodynamic method, as well as in the experimental work of processing digital stochastic signals of various objects, when it is necessary to observe the dynamics of the signal change.