Skip to Main content Skip to Navigation

Dualité entre codage de source et codage de canal

Abstract : Source coding and channel coding are dual: the former deletes redundancy while the latter adds some. Duality is therefore a key concept to simplify the complex study of joint source-channel coding, which is necessary to improve performances of transmission system under complexity constraint. In order to design such systems, we study in this PhD thesis different elementary blocks with low processing delays that rely on simple source/channel models and simple performance criteria and that use channel coding tools for source coding thanks to duality. This thesis studies the following elementary blocks: - binary source, either symmetric either asymmetric, binary symmetric channel and Hamming distance. Using linear codes or arithmetic codes, we compare a system where source and channel coding are designed separately to a system more performant called "source or channel". We give performance bounds of such a system which allows an efficient joint coding; - we use duality to compress Gaussian source under Hamming distance criterion with real BCH MDS codes. Such a coding scheme minimizes impulsive noise power; - when a Gaussian source has to be transmitted on a binary symmetric channel under mean square error criterion, we study the optimisation of source quantifier under two conditions, fixed channel code and linear relationship between source and channel dictionaries. When considering only low complexity systems, the best ones are those without channel coding! Such elementary blocks may be used in transmission schemes with unequal error protection.
Complete list of metadata
Contributor : François-Xavier Bergot Connect in order to contact the contributor
Submitted on : Saturday, February 1, 2014 - 12:04:41 AM
Last modification on : Friday, February 12, 2021 - 3:05:47 AM


  • HAL Id : pastel-00940450, version 1


François-Xavier Bergot. Dualité entre codage de source et codage de canal. Théorie de l'information [cs.IT]. Télécom ParisTech, 2000. Français. ⟨NNT : ENST 2000 E 016⟩. ⟨pastel-00940450⟩



Record views


Files downloads