Multi-Modal Semantic Communication Systems for Sustainability: A Comprehensive Survey

Abstract

The proliferation of wireless devices and the evergrowing demands for multimedia applications have resulted in bandwidth shortage and are putting a massive strain on the network resources. To overcome these shortcomings, semantic communication has come to the forefront, where only the important data features can be transmitted instead of the conventional bit-level transmission, thereby saving bandwidth and resources and achieving sustainability. Depending on the application and data, semantic features can be extracted with the help of deep learning (DL) techniques and embedded into the transmitted data. At the receiver end, the semantic features can be extracted to recover the original data or utilized to perform specific tasks. However, semantic communication research remains in its nascent stages, for which an exhaustive literature review has been presented in this domain, and the deep learning architectures employed for capturing semantics, along with the associated performance metrics, have been elaborated upon.