Research On Optimized Blind Detection And Recognition Of Communication Signals

In non-cooperative communication,the premise of identifying the signal modulation mode is to separate the signal,and the premise of the signal separation is to correctly detect the number of signal sources of the signal.Taking this as a starting point,the blind signal detection algorithm,signal separation algorithm and communication signal modulation identification algorithm are researched and optimized,and the algorithm is verified by simulation experiments and building a system experimental platform.Firstly,for the detection of the number of signal sources,the performance and application range of the detection algorithms based on the Akaike Information Ceriterion,the Minimum Description Length and the Gerschgorin Disk Estimator are studied.For the algorithm based on the Gerschgorin Disk Estimator is not limited by the noise model,this thesis improves it from the optimization of the correction factor value and the modification of the radius of the Gaelic circle.The experimental results show that the detection rate of the improved algorithm is 13.25% higher than the original algorithm.Secondly,for the problem of blind source separation,the fast independent component analysis blind separation algorithm based on negative entropy is studied and optimized.On the one hand,the joint diagonalization of the observed signal reduces the computational complexity of the algorithm.On the other hand,the number of iterations of the algorithm is reduced by using the improved Newton iteration method.The experimental results show that the improved algorithm needs fewer iterations to achieve the same separation effect,and the running time is reduced by 26.86%.Thirdly,for the modulation identification algorithm of communication signals,this thesis analyzes the difference between the high-order cumulants,instantaneous information,wavelet-transformed graphs and constellation diagrams of signals with different modulation methods,and summarizes the corresponding characteristic parameters and analyzes them.optimization.On the one hand,the eigenvalue extraction part is optimized,the wavelet transform under the optimal scale is used to optimize the extraction of the number of peaks in the graph,and a density-based clustering algorithm with flexible radius selection is used to optimize the point extraction of the constellation diagram.On the other hand,when the signals of the two different modulation modes have different characteristic parameters,the optimal characteristic parameters are selected by comparison experiments to distinguish them.Based on this,a signal modulation identification algorithm based on decision tree classifier is constructed,and experiments are carried out on 15 kinds of communication signals.The results show that the average recognition rate of the algorithm is 96.87% when the signal-to-noise ratio is5 d B.Finally,this thesis builds a system experiment platform to complete the whole process of blind detection of blind detection of the number of sources,blind signal separation and modulation mode identification,and compares the improved algorithm with the original algorithm to prove the effectiveness of the improvement.