RIS-Based on-the-Air Semantic Communications — A Diffractional Deep Neural Network Approach
Shuyi Chen, Yingzhe Hui, Yifan Qin, Yueyi Yuan, Weixiao Meng, Xuewen Luo, Hsiao‐Hwa Chen
Abstract
Semantic communication has attracted a lot of attention due to its salient features in achieving a higher transmission efficiency by focusing on semantic information delivery rather than bit-level data transmission. However, the current AI-based semantic communications rely on digital hardware for implementation. With the rapid advancement of reconfigurable intelligence surfaces (RISs), a new approach with on-the-air diffractional deep neural networks (D2NN) can be utilized to enable semantic communications in the wave domain. This article proposes a new paradigm of RIS-based on-the-air semantic communications, where the computations take place inherently as wireless signals pass through RISs. We present a system model and discuss the issues with data and control flows in this scheme, followed by a performance analysis with image transmission as an example. Compared to traditional digital hardware based approaches, RIS-based semantic communications offer many appealing characteristics, such as light-speed computation, low power consumption, and ability to handle multiple tasks simultaneously.