Preface

Tremendous traffic demands have increased in current wireless networks to accommodate for the upcoming pervasive network intelligence with a variety of advanced smart applications. In response to the ever‐increasing data rates along with stringent requirements for low latency and high reliability, it is foreseeable that available communication resources like spectrum or energy will gradually become scarce in the upcoming years. Combined with the almost insurmountable Shannon limit, these destined bottlenecks are, therefore, motivating us to hunt for bold changes in the new design of future networks, i.e., making a paradigm shift from bit‐based traditional communication to context‐based semantic communication.

The concept of semantic communication was first introduced by Weaver in his landmark paper, which explicitly categorizes communication problems into three levels, including the technical problem at the bit level, the semantic problem at the semantic level, and the effectiveness problem at the information exchange level. Nowadays, the technical problem has been thoroughly investigated in the light of classical Shannon information theory, while the evolution toward semantic communication is just beginning to take shape, with the core focus on meaning delivery rather than traditional bit transmission.

Concretely, semantic communication first refines semantic features and filters out irrelevant content by encoding the semantic information (i.e., semantic encoding) at the source, which can greatly reduce the number of required bits while preserving the original meaning. Then, the powerful semantic decoders are deployed at the destination to accurately recover the source meaning from received bits (i.e., semantic decoding), even if there are intolerable bit errors at the syntactic level. Most importantly, through further leveraging matched background knowledge with respect to the observable messages between source and destination, users can acquire efficient exchanges for the desired information with ultralow semantic ambiguity by transmitting fewer bits.

While semantic communication offers these attractive and valuable benefits, it also faces many challenges. For example, when considering the computing limitation of terminal devices, personalized background knowledge, as well as unstable wireless channel conditions, how to design semantic encoder and decoder should be a challenging issue. Moreover, in the networking layer, it is nontrivial to seek the optimal wireless resource management strategy to optimize its overall network performance in a semantics‐aware manner. Therefore, before fully enjoying the superiorities of semantic communications, this book would like to explore the following fundamental issues: (i) How many benefits can be achieved by using semantic communication? (ii) How much cost (mainly consumed resources) is incurred to guarantee the required performance, such as semantic ambiguity? and (iii) How can we maximize the benefits of semantic communications applied to different wireless networks with constraints of cost?

This book will explore recent advances in the theory and practice of semantic communication. In detail, the book covers the following aspects:

1. (1) Principles and fundamentals of semantic communication.
2. (2) Transceiver design of semantic communications.
3. (3) Resource management in semantic communication networks.
4. (4) Semantic communication applications to vertical industries and some typical communication scenarios.

Chapter 1 delves into the transceiver design for semantic communications. Specifically, we first summarize established designs for key components in semantic communications, with a key focus on the knowledge base and semantic encoders/decoders crucial for single‐user semantic communications. Our discussion then extends to multiuser SC, specifically emphasizing the synergy between various multiple‐access schemes and semantic communications, including orthogonal multiple access (OMA), space‐division multiple access (SDMA), non‐orthogonal multiple access (NOMA), rate‐splitting multiple access (RSMA), and model‐division multiple access (MDMA). In the end, we explore various applications of semantic communications and analyze the potential alterations in transceiver design required for these applications.

Chapter 2 studies semantic communications from a networking perspective, particularly focusing on the upper layer. Our primary objective is to investigate optimal wireless resource management strategies within the semantic communications‐enabled network (SC‐Net) to enhance overall network performance in a semantics‐aware manner. This entails addressing the unique challenge of ensuring background knowledge alignment between multiple mobile users (MUs) and multitier base stations (BSs). Efficient resource management remains paramount within the SC‐Net, offering numerous benefits such as guaranteeing high‐quality SemCom services and enhancing spectrum utilization. By devising effective resource allocation strategies, we aim to optimize network performance and facilitate seamless communication within the SC‐Net ecosystem.

Chapter 3 innovatively proposes the transformation theory of the semantic domain–spatial domain, projecting the knowledge graph onto a three‐dimensional tensor in the spatial domain. By mapping the entity's semantic ambiguity with the intensity of discrete points, the knowledge graph is reconstructed from background knowledge libraries and three‐dimensional tensors at the receiving end. Additionally, this chapter proposes a graph‐to‐graph semantic similarity (GGSS) metric based on graph optimal transport theory to evaluate the similarity of semantic information before and after transmission, as well as a semantic‐level image‐to‐image semantic similarity (IISS) metric that aligns with human perception. Finally, we demonstrate the effectiveness and rationality of the framework through simulations.

Chapter 4 first introduces the problem that neural networks in semantic communication are very vulnerable to adversarial attacks, then proposes robust semantic communication systems for image and speech transmission. Meanwhile, this chapter discusses the privacy issue caused by the difference of knowledge bases between transmitter and receiver in semantic communication and proposes a knowledge discrepancy‐oriented privacy protection (KDPP) method for semantic communication to reduce the risk of privacy leakage while retaining high data utility.

Chapter 5 investigates the means of knowledge learning in semantic communication with a particular focus on the utilization of Knowledge Graphs (KGs). Specifically, we first review existing efforts that combine semantic communication with knowledge learning. Subsequently, we introduce a KG‐enhanced semantic communication system, wherein the receiver is carefully calibrated to leverage knowledge from its static knowledge base for ameliorating the decoding performance. Contingent upon this framework, we further explore potential approaches that can empower the system to operate in evolving an knowledge base more effectively. Furthermore, we investigate the possibility of integration with large language models (LLMs) for data augmentation, offering additional perspective into the potential implementation means of semantic communication. Extensive numerical results demonstrate that the proposed framework yields superior performance on top of the KG‐enhanced decoding and manifests its versatility under different scenarios.

Chapter 6 presents a novel framework called VISTA (VIdeo transmission over Semantic communicaTion Approach) for video transmission by exploiting semantic communications. VISTA comprises three key modules: the semantic segmentation module and the frame interpolation module, responsible for semantic encoding and decoding, respectively, and the joint source‐channel coding (JSCC) module, designed for SNR‐adaptive wireless transmission.

Chapter 7 proposes a content‐aware robust semantic communication framework for image transmission based on generative adversarial networks (GANs). Specifically, the accurate semantics of the image are extracted by the semantic encoder and divided into two parts for different downstream tasks: regions of interest (ROI) and regions of non‐interest (RONI). By reducing the quantization accuracy of RONI, the amount of transmitted data volume is reduced significantly. During the transmission process of semantics, a signal‐to‐noise ratio (SNR) is randomly initialized, enabling the model to learn the average noise distribution. The experimental results demonstrate that by reducing the quantization level of RONI, transmitted data volume is reduced up to 60.53% compared to using globally consistent quantization while maintaining comparable performance to existing methods in downstream semantic segmentation tasks. Moreover, our model exhibits increased robustness with variable SNRs.

Chapter 8 first proposes an integrated framework for bridging meanings of semantic information in the Metaverse to achieve efficient interaction between physical and virtual worlds. This chapter then presents a Zero Knowledge Proof‐based verification mechanism to secure the authenticity of the extracted information. This chapter also introduces a diffusion model‐based resource allocation mechanism to maximize the utility of resources. Simulation results are presented to validate the authenticity and efficiency of the proposed...