Прогнозиране на амплитудни нива на шумове чрез Feed-Forward и Generalized Regression Neural Networks
Noise Amplitude Levels Prediction by Feed-Forward and Generalized Regression Neural Networks

Summary/Abstract: The report proposes a methodology for Descriptive analysis of signals with superimposed noises and training of models for predicting amplitude variations of noises using Artificial Neural Networks. The main stage is the pre-processing of Sinusoidal, Rectangular, Triangular and Triangular signals with added Uniform White Noise (UWN) and Periodic Random Noise (PRN). Through the specified procedures, Min and Max statistics are acquired for UWN, RMS and SD when processing PRN as independent predictor variables. The actual synthesis stage is associated with the formation of input data sets, training, validation and testing of Neural Networks with forward signal propagation and back-propagation of the error and Generalized Regression Neural Networks

• Details
• Contents