Fourth Cumulant Blind Source Separation Efficiency Evaluation in the Task of Cognitive Radio

Abstract

Blind source separation algorithms as a part of cognitive radio can be actively used in the task of building a smart transport system with a large number of licensed and unlicensed users. In this paper, the efficiency of the algorithm for blind separation of two complex signals is evaluated. The algorithm uses as a criterion for source separating the system of equations, which simultaneously nullify output signals covariance and fourth-order mixed cumulant. Unlike iterative methods, the considered algorithm is performed in a finite number of arithmetic operations. The paper investigates the similarity limit of input signals, separation efficiency depending on the sample size and signal-to-noise ratio. It is shown that the proposed algorithm makes it possible to effectively separate linear combinations of independent signals with a difference in the signal-to-interference ration in the input signals of at least 1 dB. It is shown that the proposed algorithm performs efficient separation of signals when the size of the sample used to calculate the statistics of the second and fourth orders is at least 10000. Compared to the FastICA algorithm, the proposed algorithm requires three times less samples to detect FSK-2 signals. It is shown that the proposed algorithm performs effective signal separation when the signal-to-noise ratio of the input signals is at least 24 dB.