Joint source-channel coding and decoding techniques are rapidly emerging as important design tools for the complexity- and latency-constrained transmission systems that underlie the multimedia revolution. In such scenarios these techniques have the potential to offer the same performance with less complexity or delay compared with systems which completely separate the source and channel coding functions. The investigators develop theoretical insights and practical approaches into the construction of joint source-channel codes by addressing certain classes of optimal variable-length error correcting codes and their application to networks. The insights gained by these studies facilitates new advanced coding strategies as well as a fundamental theory that guides the design of future wireless and multimedia systems.
One approach to joint source-channel coding is to consider prefix condition codes which have an inherent source compression property and are in addition able to detect or correct errors resulting from noise on the communication channel. Thus, these codes typically require a smaller amount of redundancy than classical channel codes to achieve the same amount of error resilience. However, the current understanding of these codes remains rather limited which has led researchers to focus on heuristics. The investigators address these issues by tackling research problems in three interrelated directions. In particular, we (1) develop new minimum-redundancy variable-length codes with inherent error resilience properties and investigate issues related to their design and compression performance; (2) study the application of this family of codes to sensor and ad hoc networks and gain insight into their repercussions for higher communication layers; (3) investigate the construction of reversible variable-length codes including those with special error-correcting properties.
