Research On Blind Source Separation Technology Of Modulated Signals

The rapid development of communication technology makes the existing spectrum resources become precious. In order to realize the effective utilization of spectrum resources, the polarization multiplexing technique has developed rapidly, and the dualpolarized antenna technology has received widespread concern for its low cost and high performance characteristics. In the non-cooperative environment, it becomes very common that vertical(or horizontal) polarized antenna receives multiple signals, which brings great challenges to signal processing and information acquisition. In this paper, the modulation recognition and blind signal separation algorithm of the aliasing signal are studied in the condition that the dual polarized antenna receiving polarization isolation greater than the emission polarization isolation. The estimation performance of the algorithms is compared by simulations.Firstly, the basic theory and implementation algorithm of cyclic spectrum are introduced, and the strip spectral correlation algorithm(SSCA) is introduced in order to improve the processing speed. Depending on the different forms of different modulated signals in cyclic spectrum cross-section, the spectral properties of cyclic spectrum are used as characteristic parameters to design the classifier, and to achieve the recognition of different modulated signals. The simulation verifies the rationality of the algorithm.Secondly, the basic concepts and properties of cyclic cumulants are analyzed, and the classifier of modulation characteristics recognition is established by using the simultaneous second-order and fourth-order cyclic cumulants. Simulations demonstrate that compared with the recognition method of cyclic spectrum, the method of cyclic cumulants has smaller amount of calculation and better real-time performance, but its recognition rate is slightly lower in the same signal-to-noise ratio.On the basis of the modulation recognition, the blind separation algorithm of the aliasing signal is studied. First of all, the single channel model is extended to multichannel model by using the interval sampling method, and the separation effect of the Fast Independent Component Analysis(Fast ICA) algorithm is analyzed. The improved Fast ICA algorithm and that based on cyclostationary constraint are introduced to further reduce the computational complexity and improve the separation performance. Simulation results show that the improved Fast ICA algorithm based on cyclostationary constraint can solve the problem of permutation order uncertainty in the traditional blind source separation method, and improve the separation accuracy and speed.Finally, in order to further improve the separation performance of the aliasing signal, on the basis of analysising the basic principle and the separation effect of the joint approximative diagonalization of eigenmatrix(JADE) algorithm, this paper adopts a joint algorithm by combining the JADE algorithm with the second-order blind identification(SOBI) algorithm together and the cyclostationary characteristic of the emission signal is applied to JADE algorithm and joint algorithm. Simulation results indicate that the joint algorithm based on the cyclostationarity optimization can achieve the blind source separation in lower SNR, and further improve the separation effect.