Modulation classification for cognitive radios with robustness against non-stationary additive noise

Modulation classification (MC) is an important field in cognitive radios. Most MCs based on the cyclic spectral density (CSD) can only apply to the stationary noise environment, but fail to inhibit non-stationary noise which is common in practical use. In this paper, we propose a MC method for cognitive radios with robustness against non-stationary additive noise, which is based on the fourth-order cyclic polyspectrum (FOCP). The proposed method is robust to non-stationary noise and can identify the modulation scheme of primary user's signal in cognitive radios. Monte Carlo simulation results demonstrate that our method can achieve better classification accuracy than the existing CSD-based MC method in non-stationary additive noise case.