This dissertation presented a first attempt at studying joint source and channel coding over fading channels. Besides refreshing the reader’s mind about well-documented theories and systems, it shows that it was possible to implement a communication system including a rather unique vocoder, a multipath, fading channel and part of a real-world mobile communication system.
While the aim of the project was apparently somewhat ambitious in comparison with its means, thus restricting the reach of its results, the author managed to put to shreds a misconception he was carrying, train his C and Matlab skills, and work on a rather interesting rate adaptation problem.
Nonetheless, the project could be pursued, maybe in the framework of a PhD thesis, in order to experiment about joint coding over a fading channel. The first step to consider, would be to integrate the finalized version of the TCENLP into the application, thus allowing to get more significant results, directly at the speech data level—such as, subjective analysis of the decoded speech, etc.
Furthermore, how to take advantage of known joint coding techniques, such as IA assignment, within the TCENLP vocoder, in order to replace the arbitrary index assignment currently taking place, could also be further studied, thus opening one axis of research. Another axis worth exploring would be to simplify the source of data, dropping the vocoder and replacing it by an abstract binary source, to concentrate only on how to adapt the generalisation of the Lloyd-Max algorithm presented in [4] from a binary symmetric channel to a fading channel.