A Study Towards Implementing Various Artificial Neural Networks for Signals Classification and Noise Detection in OFDM/PLC Channels

The presence of noise in PLC can eventually lead to information corruption. In this design, we present the usage of several classification learners in detection of noise that might found in received PLC signals at the receiving end of the OFDM channel. A database of 5,000 PLC signals with their corresponding categories was used for training and evaluation. Four neural networks were studied through experiments: radial basis function (RBF) neural network, supervised Kohonen network, counter propagation neural network, and X-Y fused neural network. The results of the experiments indicate that the RBF model achieves the best performance among the proposed methods, overall classification accuracy of 98.2%. Furthermore, the remaining proposed algorithms: CPNN and XYF networks are considerably robust classification learners, resulting in true classification percentages of 87.9%, 95.3% and 92.1% respectively.