A novel algorithm for multiple signal classification with Optimized Coulomb Energy Neural Networks for Power line communications

With the advancement in modulation schemes and cognitive techniques, Powerline Communications (PLC), have gained tremendous importance as a medium for transmission of variety of signals. The varied signals when sent through a common channel require a rigorous classification procedure for effective routing at both the transmission and receiving ends. In this paper we present complete software based, low cost, approach for classification of signals sent via the powerline. The algorithm entails structured preprocessing of the received signals, and ensemble them further for effective classification using a novel Optimized Coulomb Energy Neural Network (OCENN). The simulation and experimental results obtained shows an accuracy of more than 97% which is much better than the results of the comparative hardware approaches, which are costly and difficult to implement. It has been noticed in our experimentations that noise and attenuation experienced over the powerline affecting the higher frequency signals does not have an impact on our classification procedure, thus providing a robust architecture for implementation of PLC.