Application of Radial Basis Probability Neural Network to Signal Recognition

摘要

Radial basis probability neural network (RBPNN) was introduced to recognize radar emitter signals. The structure and training algorithm of RBPNN were firstly discussed. Then, a novel feature extraction approach called resemblance coefficient method and its detailed steps were presented. Finally, based on resemblance coefficient features, RBPNN was used to design classifer to identify 9 typical radar emitter signals. Because RBPNN inherits the advantages of both radial basis function neural network and probability neural network, RBPNN has the good characteristics of simple structure, fast learning speed and strong capabilities of pattern recognition and classification. Experimental results show that high accurate recognition rates are achieved and the introduced approach is effective and practical.