Parameter Estimation And Sorting For Multiple Frequency Hopping Signals

Two key technologies in frequency-hopping reconnaissance are studied in this paper,including parameter estimation and sorting of network stations.FH communication is widely used in military and civilian fields due to its excellent security performance,strong antiinterference ability,low probability of interception and flexible networking capability.FH communication has become an important method of anti-interference and anti-reconnaissance in wartime communication confrontation.The signal intercepted by the receiver is a mixture of signals with no prior information,and the number of receiving devices is limited in practical applications.Therefore,the study of this paper focuses on the blind parameter estimation of multi-network FH signals and the sorting of the network under underdetermined conditions.The contributions of this paper include the following three aspects:First of all,a new hopping time and hop cycle estimation algorithm based on maximum energy difference is proposed in this paper,to improve the anti-noise performance and the estimation accuracy.First,the TF matrix of the mixed signal is obtained by TF analysis.Then the carrier frequency of the signal is estimated according to the TF energy distribution,and the signal segment and the length of window are determined based on the TF data.Next,hopping time and hop cycle are estimated based on the maximum energy difference.Finally,center clustering algorithm is used to cluster the estimated values to obtain the final result.The simulation results illustrate that the new algorithm based on maximum energy difference has better anti-noise performance and estimation accuracy,and is suitable for asynchronous and synchronous networking.Secondly,a new mixing matrix estimation algorithm based on single source point detection is proposed to sort FH network stations under UBSS model.Firstly,STFT is used to obtain the sparse representation of the signal,the threshold is set to remove the TF points which are greatly affected by the noise.Then the TF points are filtered according to the single source point criterion.Next,the potential function clustering method is utilized to cluster the angles of the TF ratios at single source points.The mixing matrix estimation and the DOA estimation are obtained according to clustering centers,and finally the sorting of the network is realized.Experiments indicate that the algorithm based on single source point detection and potential function clustering method has better mixing matrix estimation performance than the comparison algorithm.Finally,an improved frequency tracking algorithm based on FFT,multi-channel clustering and autoregressive moving average model is proposed and a real-time sorting algorithm based on particle filter is introduced in this paper.Firstly,FFT is performed on the received signal samples,the number of source signals and the carrier frequency are estimated by clustering method.Then the model coefficients are estimated to construct the model,the signal is predicted and the threshold is estimated.Finally,hopping time is detected based on the error of the predicted value,and real-time frequency tracking is realized.The simulation experiment verifies the effectiveness of the proposed algorithm.The experimental results show that the proposed algorithm has higher frequency estimation accuracy than ARMA method after receiving enough signal samples,and has less computation and complexity than the Sparse Bayesian Learning method.The real-time sorting of the network station is also realized through simulation.