K-distribution fading models for Bayesian estimation of an underwater acoustic channel

Abstract

Current underwater acoustic channel estimation techniques generally apply linear MMSE estimation. This approach is optimal in a mean square error sense under the assumption that the impulse response fluctuations are well characterized by Gaussian statistics, leading to a Rayleigh distributed envelope. However, the envelope statistics of the underwater acoustic communication channel are often better modeled by the K-distribution. In this thesis, by presenting and analyzing field data to support this claim, I demonstrate the need to investigate channel estimation algorithms that exploit K-distributed fading statistics. The impact that environmental conditions and system parameters have on the resulting distribution are analyzed. In doing so, the shape parameter of the K-distribution is found to be correlated with the source-to-receiver distance, bandwidth, and wave height. Next, simulations of the scattering behavior are carried out in order to gain insight into the physical mechanism that cause these statistics to arise. Finally, MAP and MMSE based algorithms are derived assuming K-distributed fading models. The implementation of these estimation algorithms on simulated data demonstrates an improvement in performance over linear MMSE estimation.