A Novel Morphologic Filtering Algorithm for Colored-Background Noise Suppressing of Parameter Estimation

According to the theory of cyclostationary, some discrete spectral lines, located in the linear combination of signals’ carrier frequencies and symbol rates, are present in the nonlinear transform of signals to reflect cyclostationarity of communication signals. So, signals’ basic parameters can be estimated by extracting the spectral lines of nonlinear transform, but it will not only generate sinusoidal component of signal parameters, but colored-background noise will be formed as well. This phenomenon is harmful for the identification of spectral lines especially when the signal SNR(Signal-to-Noise Ratio) is low and available sampling data is not enough. Aiming at this problem, our paper delves into the frequency characteristic of nonlinear transform, employs the basic theory of mathematic morphology, and proposes a novel algorithm based on the discrete gray-scale morphologic filtering to suppress the noise. Firstly, opening operation is exploited to estimate the colored-background noise in the spectrum of signal nonlinear transform; then top-hat operation is carried out to accomplish the whiten treatment; after filling up the miscellaneous negative pulse via closing operation, the relative amplitude of spectral lines are enhanced; finally, setup detection threshold and locate the discrete spectral lines to accomplish the parameter estimation of signals. Take the estimation of QPSK symbol rate for example, all-sided Monte Carlo simulations are employed to show the relationship between detective performance and signals’ pulse shape coefficient, SNR, symbols. All the simulations take another method into account as a contrast. The results indicate that the novel algorithm suppresses colored-background noise effectively and improve the lines detection capability in the case of low SNR, small pulse shape coefficient and signal symbols. Consequently, the efficiency of the proposed algorithm is confirmed. At last, in order to take full advantage of the former spectral data, a smoothing filter operation is used to improve the detective performance in the situation of low signal SNR.