[ABT-1] Integrating Communication, Sensing, and Learning Toward the Digital-Twin Era: Semantic Communications, Representation Learning, and Optimal System Design

1:35 PM - 2:15 PM JST(4:35 AM - 5:15 AM UTC)

[ABT-1-01] Deep Joint Source–Channel Coding and Its Applications

〇Katsuya Suto¹ (1. Hokkaido University)

Keywords:

Semantic Communications,Deep Joint Source-Channel Coding,Cooperative Perception,Brain Signnal

Semantic communications have attracted significant attention as a key enabling technology for AI-native communications in Beyond 5G/6G systems by efficiently transmitting information that is meaningful to receivers. Deep Joint Source-Channel Coding (DJSCC) is one of the most promising approaches for semantic communications, where source coding, channel coding, and modulation are jointly optimized through end-to-end deep learning. Owing to its pseudo-analog transmission mechanism, DJSCC provides graceful degradation and achieves robust performance, particularly in low signal-to-noise ratio (SNR) environments. This paper reviews the fundamental principles of DJSCC from the perspectives of SoftCast and deep autoencoders, and introduces recent implementation efforts over OFDM and practical 5G systems. Furthermore, two representative applications are presented: distributed DJSCC for V2X cooperative perception and EEG-based thought-sharing communications. Finally, future research directions toward multimodal semantic communications and AI-native wireless systems are discussed.