Neuro-fuzzy networks in adaptive coding of telecommunications channels

Within the paper a system that controls the adaptive coding scheme of time varying telecommunications channels is presented. It is implemented by Berenji's neuro-fuzzy network GARIC (Generalized Approximate Reasoning-Based intelligent Control). A major advantage of this approach is that the system does not rely on any specific model of the channel. Another important advantage is that the system is able to adapt itself both to short and long term changes of the channel characteristics. A large part of the paper is devoted to the experimental work. An important conclusion is that, due to its adaptation capability, the system proves a large degree of flexibility and, in the same time, performances comparable or even better than the ones obtained using systems that "know" the model of the channel they are working on.