Pavlov

The authors propose a way to adapt the neural network algorithm based on an assessment of the average frequency of fluctuations in realization of the input signal and tuning of the clock frequency in estimating the number of samples in each interval between zeros of signal implementation.It allows you to achieve invariance of distribution of duration of intervals between zeros with respect to velocity of objects classified by the neural network.

 For solving classification problems of air and ground targets the authors justify informative signal features and robust neural network algorithms of formation of decision-making areas; these algorithms use prior information on initial statistical characteristics of acoustic signals.

 The article considers algorithms of detection and recognition of random non-centered signals and disturbances in remote strobes of pulse locators. Neural network and regression algorithms were proposed for formation of decision-making areas. For single deterministic signals quasi-optimal filtering was considered.