SUPR
Relayed Deep Joint Source-Channel Coding
Dnr:

NAISS 2025/22-1142

Type:

NAISS Small Compute

Principal Investigator:

Didrik Bergström

Affiliation:

Linköpings universitet

Start Date:

2025-08-29

End Date:

2026-03-01

Primary Classification:

20299: Other Electrical Engineering, Electronic Engineering, Information Engineering

Webpage:

Allocation

Abstract

We investigate the ability of Deep Joint Source-Channel Coding (DeepJSCC) to be relayed in a decode-and-forward setting over noisy channels and in a hybrid mobile multi-hop setting. To measure the performance we use Peak Signal to Noise Ratio (PSNR) and Learned Perceptual Image Patch Similarity (LPIPS), and we compare a baseline ("regular" DeepJSCC) and our proposed method (DeepJSCC + DHD (Deep Hash Distillation)). The motivation behind our model is to improve perceptual quality of reconstructed images.