Computation-Aware Adaptive and Scalable Deep Joint Source-Channel Coding for Heterogeneous Broadcast

Abstract

In recent years, Deep Joint Source-Channel Coding (DeepJSCC) has shown significant advantages in wireless image transmission. However, in practical broadcast scenarios, receivers often exhibit substantial heterogeneity in computational capability, bandwidth, and reconstruction requirements, which poses major challenges to existing DeepJSCC schemes that are primarily designed for point-to-point communication. To address this issue, this paper proposes a scalable DeepJSCC framework for broadcast communication, in which a single transmitter can simultaneously support receivers with diverse computational resources and bandwidth constraints. The transmitter generates multi-level codestreams, enabling receivers to intercept appropriate portions of the transmitted representation and achieve adaptive reconstruction quality without requiring multiple transmitters or separate models. To ensure reliable scalability, we further propose a knowledge distillation-based segmented training strategy, which eliminates the need for explicit feature selection and position restoration. As a result, receivers can reconstruct images by sequentially intercepting the codestream without relying on indexing information. Experimental results demonstrate that the proposed method achieves strong effectiveness and scalability under various computational configurations and codestream interception conditions.