On SNR Estimation in Time-Varying Channels Under Model Misspecification

We characterize the performance of a signal-to-noise (SNR) estimator under model misspecification. Specifically, we take the maximum-likelihood SNR estimator in the data-aided case for a slowly fading channel and apply it to a general, time-varying channel model. The saddlepoint method is used to numerically approximate the distribution of the SNR estimate in the misspecified channel and compute the probability of making a decision error in the context of outage probabilities. We discuss extensions of the analysis to broader classes of additive noise and channel gain distributions. Numerical results show close agreement between our approximations and empirically computed distribution values and decision error rates.

Duke Scholars

Author Christ D Richmond Pierre R. Lamond Department of Electrical and Computer Engin ...