A New Method for Identifying Electromagnetic Radiation Sources Using Backpropagation Neural Network

Artificial neural networks (ANN) have been widely applied for their intelligence in pattern recognition and microwave components. For an electromagnetic (EM) radiated noise source involving EM compatibility issues, however, only Missouri-Rolla University utilized the ANN to identify EM radiated noise source types based on their different frequencies. In this study, a new method was proposed to identify EM radiated noise sources by using their spatial characteristics as unique parameters for the ANN, which is available even though radiated noise source frequencies overlap. A 3-D receiver array is used to collect the spatial characteristics of radiated noise sources. By training with 60 000 datasets, the network can identify the radiated noise source type within a few minutes with an accuracy as high as 99.98%. The influences on identification accuracy from different parameters, such as ambient noise, interval between neighboring receivers, distance between receiver array and radiated noise sources, moving steps of receiver array, training data number are discussed in detail. This method can be used in EM interference diagnostics and radio monitoring with high speed and accuracy.