RECURRENT NEURAL NETWORK FOR CONTROLLING THE SPECTRUM WIDTH OF A NON-STATIONARY RANDOM SIGNAL

The purpose of the study is to develop a recurrent neural network for detecting moment beginning transient process random non-stationary signal in sliding window. possibility using apparatus artificial networks (ANN) analyzing processes investigated. Rapid detection at which changes its behavior an urgent task electrical engineering.
 
 Materials and methods. A comparison made between use autocorrelation function algorithm based on control noise.
 Results. novelty consists developing new estimating spectrum width ANN apparatus. It shown that are capable processing original signal. They do not require special pre-processing data preparation. quality operation depending parameters signals size window was carried out. ways improve architecture enrich classifier proposed.
 Conclusions. found there optimal ratio time change window, imposes restrictions choice latter method applying trained model.