Robust cyclostationary signal detection algorithm under low SNR

This paper proposes a new method to detect cyclostationary signal and estimate cyclic frequency. The proposed method exploits cyclostationary signal's recursive property to construct fourth-order autocorrelation matrix, and uses the relationship between the matrix's eigenvector and eigenvalue to achieve signal detection and cyclic frequency estimation. Classical methods realize the detection of cyclostationary signal through the estimation of cyclic autocorrelation function or cyclic spectrum. Compared with the classical methods, the proposed method exploits more prior information of cyclostationary signal, so it has a better performance of detection under low signal to noise ratio and low snapshots. The simulation results demonstrate that the pseudo cycle spectrum estimated by the proposed method is smoother than the cycle spectrum estimated by classical method, and under the same signal to noise ratio and snapshots, its detection probability of cyclostationary signal is always higher than classical method, the improvement is more obvious under low signal to noise ratio.