Research On Radar Signal Sorting And Recognition Based On Evolutionary Neural Network

In recent years,the United States has promulgated a series of new electronic warfare regulations.This move has promoted the process of shifting the focus of modern warfare operations to the electromagnetic spectrum space.Radar electronic reconnaissance is an important part of electromagnetic spectrum operations,and the recognition of radar signals directly determines the performance of the radar electronic reconnaissance system.In addition,modern warfare will occur in various environments and weather.Complex noise will affect the accuracy of radar signal identification methods.Considering severe conditions,such as sea clutter noise,atmospheric noise,and the other strong impulse noise,these noises do not obey the Gaussian distribution,so the accuracy of some recognition methods based on Gaussian noise environment will be significantly reduced.It is of practical significance and urgent needs to explore and study the recognition methods of various radar signals in different environments.Combined with warm intelligence optimization algorithm and neural network,radar recognition methods based on evolutionary neural network are designed in this paper.The main content of this paper can be divided into the following three aspects:(1)Combined quantum computation and swarm intelligence optimization algorithm,quantum swarm intelligence optimization algorithm is designed in this paper.Test functions are used to verify their performance.Original algorithm is easy to fall into local extremes or has slow rate of convergence.The experimental results show that the novel algorithms solve these problems,and the convergence accracy is improved at the same time.The designed quantum swarm intelligence optimization algorithms are used to optimize the BP neural network and the probabilistic neural network.The optimized network model has better network performance than original network model.(2)For the problem of radar signals recognition based on the intra-pulse modulation features,approximate entropy,norm entropy,harmonic mean fractal box dimension and information dimension are selected in this paper.It can be seen that they have better intra-class aggregation and inter-class separation under different signal-noise ratio(SNR)conditions in Gaussian noise environment from simulation results.The feature vector composed of these four feature parameters is used as the input of the probabilistic neural network(PNN).The quantum water evaporation optimization algorithm is designed to find the optimal smoothing parameter of the PNN.Finally,the optimized probabilistic neural network is used to identify seven kinds of radar modulation signals.The simulation results show that this method has better recognition performance and the recognition accuracy can reach 99% when the SNR is in the range of-5 d B to 20 d B.(3)The performance of traditional methods based on second-order and above-order statistics will decrease or even fail under the impact of the impulse noise which obeys alpha-stable distributed.A radar signals recognition method based on fractional lower order statistics proposed in this paper.First calculate fractional lower order statistics of intercepted radar signals.Then perform Fourier transform and other processing to obtain the intra-pulse modulation features of the radar signal.Finally identify seven kinds of radar modulation signals,and the simulation results show that when the alpha is greater than 1,the method has better recognition performance and the recognition accuracy can reach 80% in the range of generalized signal-to-noise ratio of 5d B to 20 d B.