Blind Source Separation In Application Of Signal Detection

Blind source separation(BSS) has very important theoretical and practical value in signal probing and signal processing application. Separation of sources consists of recovering a set of signals of which only instantaneous linear mixtures are observed. Given the observed data, blind sources separation aims to estimate both the separation matrix and source signals. In recent years, because of its development foreground, BSS problem has brought much attention to several research fields. Traditional BSS methods making use of independent assumption have been developed systematically. However, BSS methods are not very suitable for practical application in an increasingly complex electromagnetic environment. On the basis of separation condition of traditional BSS algorithm, this paper has explored the BSS method in statistical domain with practical context, which make use of a corresponding prior information to improve the performance of BSS algorithm. Moreover, combining with passive location, this paper has solved some hardware and algorithm problems while applying BSS methods into practical use.First, this paper will make a summary and description of separation assumption in the aspect of the requirement of derivation of BSS algorithm, and give a more intuitive explanation of separation condition by simulation result. We will make a classification of both traditional BSS algorithm and signal type in statistical domain for later derivation.Second, this paper will take advantage of the idea, adding a prior information in statistical domain, to improve the algorithm’s performance. Combining with cyclostationary signal separation algorithm, this paper proposes a new method to detect cyclostationary signal and estimates cyclic frequency at low SNR. Moreover, this paper proposes the BSS algorithms for binary phase shift keying(BPSK) and quadrature phase shift keying(QPSK) signals. The known distribution characteristics over finite points of BPSK and QPSK signals are employed to improve source separation performance. Performance analysis is conducted to derive the mean upper bound of the relative error of the separation matrix.Third, combining with the specific context of the external radiation targeting, this paper will apply the separation method to engineering practice. The works include two subjects, which are proposing new least mean square(LMS) algorithm based on function control to improve the performance of echo cancellation and implementing BSS method(AMUSE algorithm) on graphic process unit(GPU) in parallel to increase its availability.