Adaptive Channel Equalization in the Time-Varying Underwater Acoustic Channel: Performance Characterization and Robust Equalizers

摘要

Channel-estimate-based equalizers are adaptive coherent equalizers for which observations of the received signal are used to estimate channel parameters and these estimates are used to calculate the equalizer filter weights. Traditional channel-estimate- based equalizers calculate filter weights assuming that the estimates of the channel parameters are perfect. This work presents a common framework for evaluating both the performance of channel- estimate- based equalizers when the channel estimates are perfect (i.e. the minimal achievable error of the equalizer) and the degradation in performance of these equalizers due to errors in the channel estimates (i.e. the excess error). For the three type of equalizers considered (DFE Linear MMSE and Passive Time Reversal) the expressions for minimal achievable error take the form of the results from classical estimation theory for estimation error achieved by MMSE and matched filter estimators. These expressions are interpreted to give insights into the characteristics of 'good' and 'bad' channels. For the case when the channel estimates are MMSE estimates of the channel-impulse response, the excess error is shown to be proportional to the 2- norm of the calculated feedforward filter weight vector of the equalizer. This result is analogous to the 'white noise gain' result characterizing the sensitivity of adaptive array processors to mismatch. This result is used to evaluate the relative sensitivity of all three types of equalizers to environmental mismatch. The analytic predictions of equalizer performance are compared with observed performance using data from several field experiments in different underwater acoustic environments. The expressions for minimal achievable error and excess error give insights into potential methods of improving the robustness to channel mismatch of adaptive equalizers such as the DFE. Several of these methods are implemented and evaluated.