Information Embedding and Related Problems : Recent Results and Applications

Some applications and generalizations of the problem of coding with side information are developed. In the context of Gaussian broadcast channels, we show how embedding codes, based on information embedding techniques, can have significant advantages over superposition codes in several scenarios of interest, particularly when upgrading legacy digital systems. Performance limits of noisy generalizations of the coding with side information problem are also developed for finite-alphabet and Gaussian-quadratic cases. The information embedding version—writing on dirty paper wearing bad glasses—may have applications to the multiple-access channel. The source coding version, which generalizes the Wyner-Ziv problem, has applications ranging from constrained relay channels to forms of sequential estimation that arise in distributed sensor networks. For the latter, distortion-rate behavior is characterized in the Gaussian-quadratic case for a sensor pipeline.