Robust Non-Data-Aided SNR Estimation for Multilevel Constellations via Kolmogorov–Smirnov Test

A novel non-data-aided (NDA) signal-to-noise ratio (SNR) estimator for multilevel constellations based on Kolmogorov-Smirnov test is proposed in this paper. The empirical cumulative distribution function (ECDF) of certain decision statistic derived from the received signal is computed and compared with pre-stored cumulative distribution functions (CDFs) or ECDFs of reference signals with known SNRs. Then, the specific SNR, with which the pre-stored CDFs or ECDFs is the most closest to the ECDF of the received signal, is selected as the estimate. The characteristic of this estimator is to convert the estimation problem to a pattern recognition one. Extensive simulation results demonstrate that, compared with the traditional Method-of-Moment (MoM) based estimators, the proposed estimator can work properly over an extending SNR range for various multilevel constellations. With limited signal samples, the estimator offers superior estimation performance than the classic M₂M₄ estimator and newly presented M₈ estimator.