Signal Shaping for Semantic Communication Systems with A Few Message Candidates

TL;DR

This paper introduces a signal shaping technique for semantic communication systems that minimizes semantic loss using a pretrained BERT model, optimizing signal sets with few message candidates.

Contribution

It proposes a novel signal shaping method based on semantic loss minimization and an efficient projected gradient descent algorithm with proven convergence.

Findings

The proposed method outperforms existing signal shaping approaches in reducing semantic loss.

The approach effectively optimizes signal sets for systems with limited message candidates.

Simulation results validate the superiority of the method in semantic communication tasks.

Abstract

Semantic communications target to reliably convey the semantic meaning of messages. It is different from existing communication systems focusing on reliable bit transmission. To achieve the goal of semantic communications, we propose a signal shaping method by minimizing the semantic loss, which is measured by the pretrained bidirectional encoder representation from transformers (BERT) model. The signal set optimization problem for semantic communication systems with a few message candidates is investigated. We propose an efficient projected gradient descent method to solve the problem and prove its convergence. Simulation results show that the proposed method outperforms existing signal shaping methods in minimizing the semantic loss.