Decision-theoretic techniques for the development and use of side information in frequency-hop radio receivers

The erasure of unreliable symbols improves the performance of most types of error-control coding if a good method is used to decide which symbols should be erased. In this thesis, Bayesian decision theory is employed to obtain such a method for use in frequency-hop communications with Reed-Solomon coding and errors-and-erasures decoding. One erasure decision rule that is derived minimizes a bound on the probability of not decoding correctly. It makes erasure decisions for a given symbol by generating a statistic that estimates the reliability of that symbol and comparing this statistic to a fixed threshold. An advantage of this approach is that the transmission of additional symbols such as test symbols or parity symbols is unnecessary; erasure decisions can be based solely on the envelope detector outputs of a noncoherent receiver. This allows for information to be transmitted at higher data rates without sacrificing performance. Decision rules are derived for specific channels containing some combination of fading, Gaussian partial-band interference, multiple-access interference, and wideband noise. Because of the complexity of the Bayesian receivers for systems with frequency-hop multiple-access interference, it is desirable to find good approximations to the Bayesian decision rule that require lower complexity. Approximations are presented that result in a negligible loss in performance. For channels dominated by wide band noise and fading, additional Bayesian erasure decision rules are derived that perform even better than the decision rules already described. For many channel combinations, extensive performance evaluations are presented that compare the performance of the Bayesian tests with the ratio-threshold test and with errors-only decoding. The sensitivities of these tests with regard to signal-to-noise ratios and other channel parameters are also given. In nearly all cases considered, the Bayesian tests provide the best performance.U of I OnlyETDs are only available to UIUC Users without author permissio