Signal Modulation Recognition Based On ANN Optimized By DE

Along with the rapid development of modern communication, the communication system changes also rapidly. Communication signal modulation mode is more diverse and complex, so that the communication environment is more complex, the same space is more and more intensive. Research and analysis these signals, extracting the modulation parameters and determining its modulation mode, both in military and civilian fields is a valuable research topic. This paper is on the basis of previous studies, applying the DE to the signal modulation recognition based on statistical model.First of all, the signal modulation recognition method has two main ways, decision theory and statistical pattern recognition method.The parameters of statistical pattern recognition need to be extracted, observing the instantaneous amplitude, instantaneous phase and instantaneous frequency of different modulated signals after simulation, choose the feature parameters which can characterize the different types.Secondly.regard the extracted feature parameters as feature vector, study signal modulation pattern recognition problem by artificial neural network.Introduce the basic principle and characteristics,using the BP algorithm into the signal modulation recognition. This algorithm convergences slowly.and easily falling into local minimum. The genetic algorithm is used to optimize the neural network structure and connection weights. The simulation test results show that network classification time is shortened, the recognition rate is improved, but the speed of convergence is yet to be improved.Finally, introduced the new differential evolution algorithm (DE), and applied to the optimization of neural network classifier. Compared with the genetic algorithm, DE algorithm adopts real coding, the parameters need to set is less. The same as genetic algorithm, is also prone to premature convergence, local optimum. In this paper, using hybrid differential scheme and the improvement of mutation operator difference algorithm (MDE), applied to the neural network signal modulation recognition system, the simulation results show the algorithm is feasible and effective, using MDE algorithm optimized neural network signal recognition is efficiency and operation time is greatly reduced.