Juang is the Motorola Foundation Chair Professor in the Georgia Tech School of Electrical and Computer Engineering and a Georgia Research Alliance Eminent Scholar. Co-authors of the paper are Professor Geoffrey Ye Li, research associate Zhijin Qin, and Ph.D. candidate Huiqiang Xie, all associated with Li’s Intelligent Transmission and Processing Laboratory at Imperial College London.

Leveraging recent strides in deep learning and natural language processing, the paper, “Deep Learning Enabled Semantic Communication Systems,” seeks to redefine communication systems at the semantic level. The research proposes DeepSC, a deep learning-based semantic communication system designed for text transmission. Built upon the Transformer architecture, DeepSC aims to enhance system capacity and minimize semantic errors by focusing on recovering sentence meaning, as opposed to the bit- or symbol-errors seen in traditional communication methods.

Additionally, transfer learning is introduced, to ensure adaptability across diverse communication environments and expedite model training. A novel metric called sentence similarity is also initiated to accurately assess the performance of semantic communications.

In comparison to traditional communication systems that overlook semantic information exchange, DeepSC proves to be more resilient to channel variations and demonstrates superior performance, particularly in low signal-to-noise (SNR) scenarios, as evidenced by extensive simulation results.

The award-winning paper was originally published in IEEE Transactions on Signal Processing, Volume 69, in 2021. Eligibility for the IEEE SPS Best Paper Award is based on a six-year window and honors the authors of a paper of exceptional merit dealing with a subject related to the Society’s technical scope, and appearing in one of the Society’s solely owned transactions.

The team will be honored at the 2024 IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP 2024) in Seoul, Korea this April.