The main contribution of this dissertation is development of capacity approaching coding and detection schemes, as well as information theoretic characterization for several important communication systems, including relay systems, multiple-input multiple-output (MIMO) systems and storage systems.

Based on turbo and low-density parity-check (LDPC) coding principles, several coding/decoding strategies are proposed for wireless relay channels, both full-duplex relay channels which provide high information rates and half-duplex relay channels which are easy to implement are considered. An analytical tool based on average mutual information is developed to characterize the convergence behavior of different relay systems. A capacity approaching performance is shown to be achievable over both ergodic and non-ergodic fading relay channels, under both frequency-flat and frequency-selective fading scenarios.

Graph-based channel detectors are proposed for MIMO spatial multiplexing systems. Viewing the channel as a factor graph, several belief propagation (BP) based iterative detection algorithms are developed. Extensive convergence analysis, complexity comparisons and performance evaluations illustrate that, a good complexity-performance tradeoff is achieved by employing the proposed detectors.

The third research thrust is concerned with storage systems, especially bit-patterned media (BPM). Iterative decoding techniques are successfully applied to BPM recording channels. For multihead multitrack recording channels, a graph-based iterative detection scheme is proposed, which maintains a low complexity with little performance loss. The last scenario is the BPM recording channels with written-in errors and insertion/deletion errors for which information theoretical limits and several detection algorithms are developed.