Blind Source Separation Technology For Space-based AIS Mix Signals

When the signals mix at time and frequency domain,close at space,it is hard to get the targetsignal using the traditional filting and beamforming way.Blind source separation (BSS)is a veryuseful kind of technique that needs no or little prior information to separate this kind of mixsignals. Based on the requirement of separating space–based AIS (Automatic IdentificationSystem)signals, the blind source separation of co-frequency mix signals was researched. Themain content is as following:1. The research of parameter estimation for AIS mix signals. AIS signal belongs to burstsignal，the disadvantage of short data length makes it difficult to estimate the parameters, afterresearching the receiving characteristics of space-based AIS, according to the large relative-delay,a parameter estimation algorithm base on hybrid location was proposed. The hybrid location wasestimated by the difference of single signal and mixes signals at parameter, then, the non-mixpart of AIS data was used to estimate the parameter. This algorithm has low complexly but highestimation accuracy, which is suit to parameter estimation in burst mode.2. The research of separation conditions and performance bound.The first thing of BSS is todecide whether mix signals can be separated. Based on the multi-channel BSS separationcondition, using oversampling technique, the single-channel mode can be turned intomulti-channel mode, then the separation condition problem can be fixed. Based on informationtheory, the performance bound of BSS was researched, and it can provide theory support toBSS.3. The research of single-channel blind source separation. Single channel BSS is anunderdetermined problem in math, it is difficult to solve, and usually the single channel BSSalgorithm with high complexly. The sticking point of AIS real-time BSS is reducing thecomplexly. Based on Viterbi, a multi-branch joint decision-feedback detection algorithm wasproposed. The algorithm taking mixing signals as single signal at each branch by time delaydifference，and using the decision-feedback information to optimize detection property, it haslow complexly but good performance.4. The research of multi-channel blind source separation. To communication signal, thetradition algorithm will be less adaptable and more uncertainty, and the noise amplify problemwill strongly affect the separation property. As mix matrix coefficient is amplitude of the mixsignal, a constant module characteristic multi-channel BSS algorithm for mix AIS signals wasgiven, this algorithm has no uncertainty problem but low complexity; it is also adapt to mix AISsignals separation. To solve the he noise amplify problem, base on the noise relativity of separated signals,a noise variance minimization method was given, this method can have abetter performance after noise reduction.5. The research of blind source separation combine multi-channel with single-channel. Thesingle-channel algorithm can make full use of the information in parameter difference, but whenthe parameter difference is small, the single-channel algorithm will lost its advantage; thesubstance of tradition BSS is maximize SIR(signal to interference ratio),when the conditionnumber of mix matrix is large, the separated signals will be lost in noise. As multi-channel BSScan change the SIR, a non-maximize SIR algorithm was proposed, this algorithm use maximizechannel capacity criteria to choose the right SIR, then use the single-channel algorithm toseparate. This algorithm not only consider the noise amplify problem, but also can make use ofparameter difference information, combine with the advantage of multi-channel withsingle-channel to have a better performance.