Joint Source-Channel Coding for Two-Hop Channels in Multihop Semantic Communication: Image Transmission Study ​

Joint source-channel coding is a semantic communication technique that preserves the quality of multimedia data (e.g., an image) by protecting it analogically from channel noise. Its effectiveness has been proven in various wireless scenarios, and its most interesting feature is its adaptability to non-trivial channels and different types of data. The most common technique to perform this is to use autoencoders, which learn the features of an image and try to reconstruct it optimally from a reduced set of features. However, most studies on joint source-channel coding involve a single link: adding a second hop, or multiple wireless hops (as in, for example, a drone network), which changes the channel scenario and opens up new possibilities. In this project, different solutions for two-hop channels will be studied, such as decoding and re-encoding and joint coding with and without re-amplification.