A fast convergence blind source separation algorithm based on cyclostationary statistics

This paper presented a new blind source separation algorithm using the cyclostationary statistics of source signals. On the basis of the blind source separation model, we first determined several useful assumptions. Next we get an updating equation of separation matrix, making use of the independence and cyclostationary statistics of source signals. Later we discussed the situations of real signals, complex signals and normalized algorithms. Finally, the paper compared the algorithm with other three algorithms and analyzed the convergence and performance index (PI). The simulation results demonstrated that the new presented algorithm had faster convergence speed, better separation effects and robustness, comparing to the other algorithms. Its convergence speed is nearly twice as fast as the speed of normalized EASI algorithm.