Research On Blind Source Separation And Its Application In Communication Reconnaissance

With the development of communication technology, the electromagnetic environment becomes all-time complicated and dense. The communication frequency band is continuously broadening. Also, the communication system and signal mode are continuously renewing. For non-cooperative receiving communication reconnaissance, it’s quite difficult to intercept, analyze and identify the threaten signals and also to determine the jamming target in so complex environment. Facing the challenges of communication countermeasures, it’s necessary to research a novel reconnaissance technology in communication countermeasures.In the practical military communication, the intercepted signals are usually the mixture of the signals such as enemy interference signals, enemy communication signals, our communication signals, various electromagnetic interference and noise signals, etc. Therefore, for reconnaissance of enemy communication signals, the first thing that we need to do is to separate the required enemy communication signals from the intercepted mixture.Due to the non-cooperation of the communication reconnaissance and the variety of the modulation mode of communication signal, generally, it’s impossible to obtain any information about the enemy transmitting signals. In addition, because of the complexity of electromagnetic environment, if we want to separate the signals from the mixture without any prior information of enemy signals, communication reconnaissance face the tremendous challenge. Aiming at the problem of communication reconnaissance facing in complex electromagnetic circumstance, we present a novel system of communication reconnaissance based on blind source separation. We want to find a new approach to solve this technical difficult problem from blind signal processing.The major achievements and results are outlined as follows.1.Linear instantaneous mixed BSS technology is formulated, which is the basis of later research work. According to linear instantaneous mixed system model, the ambiguity inherent in BSS technology and fundamental assumption are analyzed. Also, some common methods of BSS are introduced.And then,performance index of BSS is presented,for instance, interference-to-signal ratio (ISR) and interconnect error. Finally, whitening algorithm that is essential in BSS technology is deduced.2.Linear instantaneous mixed BSS algorithms are researched.Firstly, some existing BSS algorithms are researched and the derivation of these algorithms is given. In order to improve the convergence speed of the algorithms and keep stability, some BSS algorithms based on adaptive step size are put forward, such as variable step size natural gradient BSS algorithms, variable step size EASI BSS algorithms, and so on. And the stable condition of these algorithms is analyzed. The computer simulation results validate the advantage of the improved algorithms.3.Some blind separation technologies of communication signals are researched.Firstly, in order to solve the problem of the phase ambiguity in BSS of communication complex-valued signals, we present a BSS algorithm of non-phase ambiguity in communication complex-valued signals. The algorithm can remove phase ambiguity of digital communication signals which have independent in-phase and quadrature components. In addition, the algorithms do not need to set iterative step size while they have the fast convergence speed and strong stability. The simulation results show that the algorithm can separate not only the signals with the same carrier frequencies but also ones with different carrier frequencies. The results also show that the algorithm can separate the digital communication signals whether they have the same modulation modes or not, and the signals can be separated even if they have the same carrier frequencies and same modulation modes.Secondly, using the vandermonde characteristic of uniform linear array, a BSS algorithm based on DOA estimation is proposed, which can further improve the separation efficiency. Simulation results show the average ISR of the DOA-EASI algorithm is lower 7.5dB than that of EASI algorithm and the average ISR of the DOA-FastICA algorithm is lower 4.3 dB than that of FastICA algorithm.Thirdly, in order to improve anti-noise performance of BSS, on the basis of EASI algorithm,a robust EASI BSS algorithm is presented, which can weaken the effect of the noise according to noise power spectral density model.The simulation results show the robust EASI algorithm needs 5dB ISR smaller than the EASI algorithm does to obtain the same output ISR.Finally, a BSS algorithm in multi-path channel circumstance is studied, which is a convolutive mixing BSS technology. A time-domain blind deconvolution algorithm based on subspace decomposition is proposed. On the basis of the subspace decomposition, the convolutive mixing model is converted into linear instantaneous mixing model,and then the required signals are separated by the linear instantaneous mixing BSS algorithm.The algorithm does not need to solve the order ambiguity and amplitude ambiguity that need solving in frequency domain blind deconvolution algorithm.At the same time, the source signals need not be identified and independent and can be processed in IF.4.The application of BSS in communication reconnaissance is studied.Firstly, the application of BSS technology in DS-CDMA system is studied. An algorithm for pseudo-code estimation of DS-CDMA signal and blind multi-user separation technology in DS-CDMA based on BSS are proposed. Secondly, the application of BSS technology in frequency-hopping (FH) communication is studied. We suggest a new idea that separates the FH signals by using BSS technology for the first time. The simulation results prove the method validity. Thirdly, by studying the application of blind source separation in communication anti-interference technology, a communication anti-interference technology based on BSS is put forward. The simulation results show the technology can resist the interference like wideband interference with same carrier frequency, narrowband interference with same carrier frequency, and the interference with same carrier frequency and same modulation, and so on.5.A novel system of communication reconnaissance technology based on BSS is presented. The component of the communication reconnaissance system based on BSS is proposed and the major thought of the reconnaissance technology of feature matching is expatiated. And on the basis of research on the time-frequency analysis and neural network, the blind identification technology of the communication signals is proposed. First, time-frequency analysis and singular value decomposition (SVD) algorithm are used to implement data compression and the extraction of signal feature vector is realized. Then the classification and recognition to it are implemented by utilizing the learning and classification ability of neural network.6.The experiment of BSS is performed. An experimental system is established and we collect some data with this system.These data is blind separated by the BSS algorithm researched in this thesis. The results prove that BSS is effective for communication signal blind separation. At the same time, we study the parameters’ effect on the performance of BSS algorithm,such as channel interval,signal-to-noise ratio, bandwidth and spatial interval.