USING GENERAL REGRESSION NEURAL NETWORK FOR SIGNAL RESTORATION

Abstract

The objective of this study is to evaluate the potentiality of using Artificial Neural Networks. The GRNN Model was trained with 123 learning patterns. Training patterns have been generated artificially, where Work Bench Simulator software was used to produce 123 electrical signals. The signals were randomly distorted. The learning patterns were generated by attaching the variables of the original signals with the corresponding distorted ones. The model was trained for one second. A minimum / error of +0.0126 ×10-8 and smoothing factor of 0.201560 were obtained. The trained model was applied to a new set of data (25 signals). The model was capable to process new data with an error of +0.0126 ×10-8 The output results were subjected to statistical analysis. A general standard error of +6.185×10-7 was obtained. The analysis proved that the GRNN can be used for signal restoration based on good previous experience of learning.