An investigation of SNR estimation techniques based on uniform Cramer-Rao lower bound

This paper is concerned with a comparison of the performance characteristics of the signal-to-noise ratio (SNR) estimation techniques based on maximum likelihood, lower order moments, or higher order moments. This is based on a modified and more general bias gradient estimator. The uniform Cramer-Rao lower bound (UCRLB) is used to specify the achievable and unachievable regions in the bias gradient length-variance (δ - σ) trade-off plane. The UCRLB bound is also used as a benchmark for performance comparisons for biased SNR estimators. The identification of the best SNR estimator is facilitated through plotting the performance of various estimators on the same bias gradient-variance plane as obtained through Monte Carlo simulations.