Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Rateless Stochastic Coding for Delay-Constrained Semantic Communication

Cheng Peng, Rulong Wang, Yong Xiao

TL;DR

A novel JSCC coding scheme called rateless stochastic coding (RSC) is proposed by introducing a generative decoder and dithered quantization that enables reconstruction based on both distortion and perception metrics through rateless transmission.

Abstract

We consider the problem of joint source-channel coding for semantic communication from a rateless perspective, the purpose of which is to settle the balance between reliability (distortion/perception) and effectiveness (rate) of transmission over uncertain channels. In particular, we propose a more general communication objective that minimizes the perceptual distance by incorporating a semantic-level reconstruction objective in addition to the conventional pixel-level reconstruction objective. Based on the proposed objective, we then propose a novel JSCC coding scheme called rateless stochastic coding (RSC) by introducing a generative decoder and dithered quantization. The coding scheme enables reconstruction based on both distortion and perception metrics through rateless transmission. Extensive experiments demonstrate that the proposed RSC can achieve variable rates of transmission maintaining an excellent trade-off between distortion and perception.

Rateless Stochastic Coding for Delay-Constrained Semantic Communication

TL;DR

A novel JSCC coding scheme called rateless stochastic coding (RSC) is proposed by introducing a generative decoder and dithered quantization that enables reconstruction based on both distortion and perception metrics through rateless transmission.

Abstract

We consider the problem of joint source-channel coding for semantic communication from a rateless perspective, the purpose of which is to settle the balance between reliability (distortion/perception) and effectiveness (rate) of transmission over uncertain channels. In particular, we propose a more general communication objective that minimizes the perceptual distance by incorporating a semantic-level reconstruction objective in addition to the conventional pixel-level reconstruction objective. Based on the proposed objective, we then propose a novel JSCC coding scheme called rateless stochastic coding (RSC) by introducing a generative decoder and dithered quantization. The coding scheme enables reconstruction based on both distortion and perception metrics through rateless transmission. Extensive experiments demonstrate that the proposed RSC can achieve variable rates of transmission maintaining an excellent trade-off between distortion and perception.
Paper Structure (16 sections, 8 equations, 5 figures)

This paper contains 16 sections, 8 equations, 5 figures.

Table of Contents

  1. Introduction
  2. Related work
  3. Divergence-based Perception
  4. Rateless code
  5. System Model and Problem Formulation
  6. System Model
  7. Problem Formulation
  8. RSC For Delay-constrained Communication
  9. Design of RSC
  10. Dithered Quantization
  11. Optimization of RSC
  12. Experimental Results
  13. Simulation Setup
  14. Quantitative Analysis
  15. Qualitative Comparison
  16. ...and 1 more sections

Figures (5)

  • Figure 1: Illustration of a semantic communication system with CR over a noisy communication channel.
  • Figure 2: RSC method architecture diagram.
  • Figure 3: The rate-distortion-perception trade-off of MINIST images based on the RSC. (a) Equi-rate lines plotted on the R-D-P trade-off surface highlight the tradeoff between distortion and perceptual quality at different rates. (b) Cross sections of the R-D-P trade-off surface along perception-distortion planes. (c) Cross-sections of the R-D-P trade-off surface along rate-perception planes
  • Figure 4: Performance comparison of RSC with baseline JSCC schemes on the MINIST dataset over AWGN channels.
  • Figure 5: Examples of reconstructed images from baselines and the proposed RSC on MNIST and CIFAR-10 varying the survivor rate. The top row of each reconstruction snapshot is the original image, and subsequent rows are reconstructed images for a reduced rate = {12, 10, 8, 6, 4, 2}.