Structured channel methods for wireless communications.

摘要

This dissertation addresses the problem of channel estimation and data detection in a wireless channel. By exploiting prior known information about the transmit and receiver filters in linear modulation schemes, we are able to show that the overall channel impulse response vector lies approximately in the subspace of matrices associated with the transmit and receive pulse shapes. The resulting structured channel model typically has fewer unknown parameters than conventional channel models and the unknown parameters are related directly to the physical propagation channel. Acquisition and tracking of the overall channel are carried out through the structured channel vector. The smaller number of parameters also lead to smaller amount of training overhead required for channel acquisition.; After a brief introduction to the characteristics of wireless channels, we develop the structured channel models for time-invariant and time-varying channels. The structured channel estimators are then used in maximum likelihood sequence estimation (MLSE) receivers in slow and fast fading channels. We also extend the MLSE receiver algorithm to deal with CCI in dynamic channel environments. The performance of these receivers is analyzed using the pairwise error probabilities, and upper bounds on the error rates of the receivers are derived. Robustness and complexity issues of the receivers are also discussed. We follow this with the problem of designing direct minimum mean square error (MMSE) equalizers that exploit the underlying channel structure. Blind identifiability conditions for the equalizers are also derived. In certain wireless applications where no training sequences are available, blind channel identification and equalization is required to maintain reliable system performance. We shall discuss some prior work in blind channel identification and equalization and then propose blind channel estimation algorithms that exploit the underlying channel structure. To analyze the performance of blind channel estimation techniques, a general Cramer-Rao bound (CRB) under differentiable constraints on the unknown parameters is derived. This constrained CRB extends previously known results in that it does not require that the original Fisher Information matrix of the unconstrained problem be of full rank.