Joint Channel Sounding and Source-Channel Coding for MIMO-OFDM Systems: Deep Unified Encoding and Parallel Flow-Matching Decoding

In this work, we propose a deep unified (DU) encoder that embeds source information in a codeword that contains sufficient redundancy to handle both channel and source uncertainties, without enforcing an explicit pilot-data separation. At the receiver, we design a parallel flow-matching (PFM) decoder that leverages flow-based generative priors to jointly estimate the channel and the source, yielding much more efficient inference than the existing diffusion-based approaches. To benchmark performance limits, we derive the Bayesian Cramér-Rao bound (BCRB) for the joint channel and source estimation problem. Extensive simulations over block-fading MIMO-OFDM channels demonstrate that the proposed DU-PFM approach drastically outperforms the state-of-the-art methods in both channel estimation accuracy and source reconstruction quality.

💡 Research Summary

This paper addresses the long‑standing challenge of jointly performing channel sounding and source‑channel coding in modern MIMO‑OFDM systems without allocating dedicated pilot resources. Conventional deep joint source‑channel coding (DJSCC) schemes either assume perfect channel state information (CSI) or rely on explicit pilot symbols, which in current 5G‑NR and Wi‑Fi 7 deployments consume 15 %–20 % of the time‑frequency grid. The authors propose a unified solution consisting of a Deep Unified (DU) encoder and a Parallel Flow‑Matching (PFM) decoder that together eliminate the need for separate pilots while preserving high spectral efficiency.

The DU encoder maps each transmitter’s source data Sₖ into a complex‑valued OFDM tensor Xₖ using a Swin‑Transformer backbone. The encoder is trained end‑to‑end with a power constraint ‖Xₖ‖F² ≤ N_f T_s P, but without any channel knowledge at the transmitter. To facilitate training, an auxiliary decoder g{βₖ} is introduced, and the encoder‑decoder pair is optimized to minimize the reconstruction loss ‖Sₖ − g_{βₖ}(f_{γₖ}(Sₖ))‖_F². This forces the encoder to embed sufficient redundancy—effectively an implicit pilot—directly into the transmitted symbols, while still preserving the semantic content of the source.

On the receiver side, the PFM decoder leverages optimal‑transport (OT) flow models to learn generative priors for both the channel H and the source symbols Sₖ. A flow defines a continuous probability path p(H(τ)) and p(Sₖ(τ)) for τ∈