Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

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

Hao Jiang, Xiaojun Yuan, Qinghua Guo

TL;DR

This paper tackles the problem of joint channel sounding and source-channel coding for MIMO-OFDM systems by introducing a deep unified (DU) encoder that embeds source data with redundancy directly into channel inputs, eliminating the need for explicit pilot-data separation. It pairs the DU encoder with a parallel flow-matching (PFM) decoder that uses flow-based priors to jointly estimate the MIMO channel and reconstruct the source, enabling efficient Bayesian inference under limited time-frequency resources. The authors derive a Bayesian Cramér-Rao bound (BCRB) by incorporating prior information via optimal-transport flow priors, and they validate the approach with extensive simulations in block-fading MIMO-OFDM channels, showing substantial improvements in NMSE for the channel, NMSE for the transmitted signal, and perceptual quality of recovered sources, all at significantly reduced channel bandwidth ratios. The proposed DU-PFM framework reduces pilot overhead and inference time while delivering superior spectral efficiency and reconstruction quality, highlighting its practical impact for next-generation wireless systems where joint sensing and communication are advantageous.

Abstract

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.

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

TL;DR

This paper tackles the problem of joint channel sounding and source-channel coding for MIMO-OFDM systems by introducing a deep unified (DU) encoder that embeds source data with redundancy directly into channel inputs, eliminating the need for explicit pilot-data separation. It pairs the DU encoder with a parallel flow-matching (PFM) decoder that uses flow-based priors to jointly estimate the MIMO channel and reconstruct the source, enabling efficient Bayesian inference under limited time-frequency resources. The authors derive a Bayesian Cramér-Rao bound (BCRB) by incorporating prior information via optimal-transport flow priors, and they validate the approach with extensive simulations in block-fading MIMO-OFDM channels, showing substantial improvements in NMSE for the channel, NMSE for the transmitted signal, and perceptual quality of recovered sources, all at significantly reduced channel bandwidth ratios. The proposed DU-PFM framework reduces pilot overhead and inference time while delivering superior spectral efficiency and reconstruction quality, highlighting its practical impact for next-generation wireless systems where joint sensing and communication are advantageous.

Abstract

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.
Paper Structure (15 sections, 1 theorem, 16 equations, 3 figures, 1 table, 1 algorithm)

This paper contains 15 sections, 1 theorem, 16 equations, 3 figures, 1 table, 1 algorithm.

Table of Contents

  1. Introduction
  2. System Model and Problem Formulation
  3. System Model
  4. Problem Description
  5. Transceiver Design
  6. Transmitter Design
  7. Receiver Design
  8. Performance Analysis
  9. Numerical Results
  10. Performance Comparison with Idealized Transmission Schemes
  11. Performance Comparison with Classical Schemes
  12. Performance Comparison with DNN-based Schemes
  13. Conclusions
  14. Proof of Proposition \ref{['prop:FIM_rank_deficiency']}
  15. Derivation of the errors between the $\epsilon$-smoothed scores and the true prior scores

Key Result

Proposition 1

The FIM $\boldsymbol{F}(\boldsymbol{x}, \boldsymbol{h}) \in\mathbb{C}^{N_fN_t(T+N_r)\times N_fN_t(T+N_r)}$ is always rank-deficient with $\operatorname{rank}(\boldsymbol{F}(\boldsymbol{x}, \boldsymbol{h})) \leq N_fN_t(T+N_r) - N_fN_t^2$.

Figures (3)

  • Figure 1: The proposed transceiver design, where $\{g_{\boldsymbol{\beta}_k}\}_{k=1}^{N_t}$ are the auxiliary DU decoders to facilitate the training of the DU encoders $\{f_{\boldsymbol{\gamma}_k}\}_{k=1}^{N_t}$, and $g_{\boldsymbol{\varTheta}}$ is the desired PFM decoder that achieves joint channel and source estimation.
  • Figure 2: $\text{NMSE}_{\mathcal{H}}$, $\text{NMSE}_{\mathcal{X}}$, and $\text{PSNR}_{\mathcal{S}}$ of the DU-PFM framework under various CBRs and CSNRs. The pilot ratio for OP/SP-DU-PFM is $\alpha=0.5$.
  • Figure 3: Tradeoff between the estimation accuracy and the computational complexity in terms of NFE, where CSNR $=10$ dB and CBR $=1/192$.

Theorems & Definitions (4)

  • Remark 1
  • Proposition 1: Rank Deficiency of $\boldsymbol{F}(\boldsymbol{x}, \boldsymbol{h})$
  • proof
  • Remark 2: Interpretation of Rank Deficiency