SNR Estimation for Multilevel Constellations Using Higher-Order Moments

Abstract

The performance of existing moments-based non-data-aided (NDA) estimators of signal-to-noise ratio (SNR) in digital communication systems substantially degrades with multilevel constellations. We propose a novel moments-based approach that is amenable to practical implementation and significantly improves on previous estimators of this class. This approach is based on a linear combination of ratios of certain even-order moments, which allow the derivation of NDA SNR estimators without requiring memory-costly lookup tables. The weights of the linear combination can be tuned according to the constellation and the SNR operation range. As particular case we develop an eighth-order statistics (EOS)-based estimator, showing in detail the statistical analysis that leads to the weight optimization procedure. The EOS-based estimators yield improved performance for multilevel constellations, especially for those with two and three amplitude levels. Monte Carlo simulations validate the new approach in a wide SNR range.