Abstract:

In view of the deficiencies, slow convergence and false saturation phenomenon, which are present in signal recognition of the existing multilayer perception neural network classifier based on back-propagation algorithm, the combined feature module selected by a bee colony algorithm is used, and three different algorithms, quick prop, super adapt error back-propagation and conjugate gradient, are presented and used in the multilayer perception neural network classifier to realize the automatic recognition of communication signals in this paper. There proposed algorithms have a higher recognition rate compard with the error back-propagation algorithm. The simulation results show that the proposed algorithms can overcome the shortcomings of the error backpropagation algorithm, and the recognition rates are higher than 95% under the conditions that the number of neurons is only 20 in the hidden layer, the signal-to-noise ratio of 4 dB, and the system is easy to realization, and has wide application prospects in signal recognition.