Research On Anti-interference Method Of Frequency Hopping Communication System Based On Blind Source Separation

With the deepening of informatization,the quality of military communication relies more on the method of interference and anti-interference.Frequency hopping communication has become one of the most commonly used anti-interference methods.As another kind of anti-interference method,blind source separation(BSS)can use the received signal to recover the source signal without any prior information of the sending end after a certain processing.The interference is separated.Therefore,it has a good antiinterference effect.In this thesis,the two anti-interference methods are combined,the following contents are included: blind source separation of frequency hopping communication under positive conditions,mixed matrix estimation of frequency hopping communication under underdetermined conditions,and source signal recovery under underdetermined conditions.Firstly,the separation under positive definite conditions is studied.It means that the number of observed signals is the same as the transmitted signals.This thesis mainly studies the mixing of frequency hopping signals and interference commonly used in frequency hopping communication.The FastICA algorithm based on negative entropy and algorithm based on fourth-order cumulant matrix joint diagonalization are adopted.It use these algorithms to analyze the separation performance under various conditions.In order to study the influence of BSS on anti-jamming capability of frequency hopping communication system,the bit error of frequency hopping communication system when de-hopping and demodulation directly is compared to that when de-hopping and demodulation after the BSS.It's verified that using BSS before de-hopping and demodulation can reduce the bit error of the frequency hopping communication system.Then,the underdetermined blind source separation is studied.It means that the number of observed signals is smaller than that of transmitted signals.In this case,the original mixed matrix cannot be directly estimated and the transmitted signals cannot be restored easily as in positive definite condition.So,it always needs two steps to complete the separation,that is estimating the mixing matrix first and then restoring the transmitted signal.For the estimation of the mixed matrix,using the method of selecting the timefrequency single source point to estimate the mixed matrix.First,using short-time Fourier transform or other time-frequency transform methods to transform the signals into the time-frequency domain.Then to select the time-frequency signal source point in timefrequency domain.After that,cluster the selected points to obtain the mixed matrix.In addition,a method based on calculating the tangent of the signal for matrix estimation is proposed.It is verified that the above methods can estimate the mixed matrix.For the recovery of source signal,this paper adopts the shortest path method,the method based on signal statistical sparsity characteristics and the method based on subspace projection to recover the source signals.For the above several algorithms,using the simulation to verifies the feasibility of the them and obtains relevant conclusions.