Research On Recognition Algorithm Of Frequency Hopping Signal Modulation Mode In Multipath Channel

As a spread spectrum method,frequency hopping communication technology has the advantage of anti-interference.It is widely used in modern warfare radio communication and electronic countermeasures.Its main feature is the pseudo-random hopping of the signal carrier frequency,which is also the main difference with the conventional fixed frequency communication signal.The prior information required in the process of frequency hopping signal de-hopping is accurate frequency estimation,which requires time-frequency analysis to estimate the technical index parameters of the frequency hopping signal,and more accurate identification of the modulation method of the frequency hopping signal.However,the frequency hopping system will still be affected by multipath interference,especially the slow frequency hopping system.Multipath interference will cause inter-symbol interference,which will affect the accuracy of modulation recognition.This thesis mainly studies the frequency hopping signal modulation recognition algorithm based on multi-feature parameters and support vector machine(SVM).Based on the fact that the number of symbols carried by each carrier frequency is very small(Because the symbols of the frequency hopping signal are scattered on multiple carrier frequencies for transmission),and the deviation of the extracted eigenvalues,analyzed the distribution difference of the three types of eigenvalues in the selected signal set: the characteristic entropy of the wavelet energy spectrum,the correlation coefficient of the cyclic spectrum section,and the high-order cumulant.Moreover,compared the distribution difference of the eigenvalues extracted in the case of large samples and small samples,and screen out the eigenvalues suitable for the recognition of small sample frequency hopping slice signal modulation mode.Last,starting from the identification environment of the multipath channel,filter the eigenvalues that are less affected by multipath interference.Aiming at the problem that too many feature values extracted will cause the low efficiency of classification,simplified the classifier structure and improve the recognition rate of modulation recognition algorithms by constructing the two-dimensional feature parameters through combining the eigenvalues with similar classification capabilities for a specific signal set,and introducing the SVM,and by the kernel function,the two-dimensional feature distribution point set is mapped to the high-dimensional space to find the optimal classification surface,and the optimization results of polynomial,radial basis,and sigmod are compared.Aiming at the problem of the low recognition rate of some signal sets in the low signal-to-noise ratio and multipath channel environment of the classifier based on SVM construction,the paper proposes an improved SVM classifier based on the decision-level fusion model.For hierarchical support vector machines model,by setting up multiple SVM classifiers constructed with eigenvalues with similar classification capabilities,and by using the prior probability matrix to reduce the small sample deviation rate,using the form of voting judgment to synthesize multiple recognition results to determine the decision value,in the decision center.Compared the feature recognition algorithm based on fuzzy function theory,the simulation results show that by expanding the signal set,adding multipath interference,and combining with the decision-level fusion model for the algorithm in this paper,the recognition rate of BPSK,QPSK and MSK signals can reach about 90% at a signal-to-noise ratio of-5 ～ 5d B,which means that,it is more robust in low signal-to-noise ratio and multipath channel environment,and has a wider application space.