Local mean signal estimation over Nakagami‐m fading channels

Abstract

Accurate and fast algorithms for the local mean signal level estimation are very important for the successful implementation of many wireless communication enabling techniques such as handoff, power control, and adaptive gain control schemes. In this paper, we present on-line simple-to-implement (i) minimum variance unbiased and (ii) maximum likelihood estimators for the local mean signal power estimation in the decibel domain. We consider a generalized Nakagami-m fading environment and show that these estimators are asymptotically efficient in the number of samples and in the fading parameter m. Numerical and simulation results confirm that these estimators outperform the sample mean estimator and quickly approach the Cramer-Rao lower bounds.

SNR estimation in decibel domain over communication channels

“…In [4], the optimum minimum variance unbiased estimators (OMVU) have been derived under the assumption of (i) independent and identically distributed samples to be used for the estimation, (ii) Rayleigh fading, (iii) a single path. In [8], the results in [4] were generalized to Nakagami-m fading channel model. The assumptions in [8] are the same as the ones in [4] except the second assumption.…”

“…In [8], the results in [4] were generalized to Nakagami-m fading channel model. The assumptions in [8] are the same as the ones in [4] except the second assumption.…”

“…The estimators presented in [4] and [8] have the following drawbacks. Firstly, signal samples are generally correlated in most practical applications where the correlation is a function of the mobile speed.…”

“…Secondly, those two estimators are model-based, that is, they are optimum only and only if the channel is Rayleigh fading for [4] or Nakagami-m fading for [8]. As pointed out in [8], if the OMVU obtained from the Rayleigh channel model is applied to the Nakagami channel model, the performance loss of OMVU estimator is significant and even much worse than the simple sample mean estimator. Thus, if the channel does not follow those assumed fading models, the estimator [4] and [8] may cause large estimation errors.…”

“…As pointed out in [8], if the OMVU obtained from the Rayleigh channel model is applied to the Nakagami channel model, the performance loss of OMVU estimator is significant and even much worse than the simple sample mean estimator. Thus, if the channel does not follow those assumed fading models, the estimator [4] and [8] may cause large estimation errors. Thirdly, the estimator of [8] requires a good estimation of Nakagami parameter m, which is strongly model-based.…”