Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Learning-Based Signal Recovery in Nonlinear Systems with Spectrally Separated Interference

Jayadev Joy, Sundeep Rangan

TL;DR

This work tackles the challenge of robust signal recovery for wideband FR3 receivers operating under strong adjacent-band interference and front-end nonlinearities that induce spectral leakage. It introduces a learned ML-VAMP framework that preserves the interpretable, iterative structure of model-based inference while replacing intractable nonlinear denoisers with compact neural networks guided by spectral priors. The method demonstrates significant gains over linear baselines, especially in saturation-dominated, high-interference regimes, and shows resilience to quantization with a two-network, iteration-wise denoising strategy. The results suggest that hybrid inference, combining spectral filtering and learned denoisers, can enable cost-efficient, robust reception in next-generation FR3 systems, with future work extending to time-varying interference and memory effects.

Abstract

Upper Mid-Band (FR3, 7-24 GHz) receivers for 6G must operate over wide bandwidths in dense spectral environments, making them particularly vulnerable to strong adjacent-band interference and front-end nonlinearities. While conventional linear receivers can suppress spectrally separated interferers under ideal hardware assumptions, receiver saturation and finite-resolution quantization cause nonlinear spectral leakage that severely degrades performance in practical wideband radios. We study the recovery of a desired signal from nonlinear receiver observations corrupted by a high-power out-of-band interferer. The receiver front-end is modeled as a smooth, memoryless nonlinearity followed by additive noise and optional quantization. To mitigate these nonlinear and quantization-induced distortions, we propose a learned multi-layer Vector Approximate Message Passing (LMLVAMP) algorithm that incorporates spectral priors with neural network based denoising. Simulation results demonstrate significant performance gains over conventional methods, particularly in high-interference regimes representative of FR3 coexistence scenarios.

Learning-Based Signal Recovery in Nonlinear Systems with Spectrally Separated Interference

TL;DR

This work tackles the challenge of robust signal recovery for wideband FR3 receivers operating under strong adjacent-band interference and front-end nonlinearities that induce spectral leakage. It introduces a learned ML-VAMP framework that preserves the interpretable, iterative structure of model-based inference while replacing intractable nonlinear denoisers with compact neural networks guided by spectral priors. The method demonstrates significant gains over linear baselines, especially in saturation-dominated, high-interference regimes, and shows resilience to quantization with a two-network, iteration-wise denoising strategy. The results suggest that hybrid inference, combining spectral filtering and learned denoisers, can enable cost-efficient, robust reception in next-generation FR3 systems, with future work extending to time-varying interference and memory effects.

Abstract

Upper Mid-Band (FR3, 7-24 GHz) receivers for 6G must operate over wide bandwidths in dense spectral environments, making them particularly vulnerable to strong adjacent-band interference and front-end nonlinearities. While conventional linear receivers can suppress spectrally separated interferers under ideal hardware assumptions, receiver saturation and finite-resolution quantization cause nonlinear spectral leakage that severely degrades performance in practical wideband radios. We study the recovery of a desired signal from nonlinear receiver observations corrupted by a high-power out-of-band interferer. The receiver front-end is modeled as a smooth, memoryless nonlinearity followed by additive noise and optional quantization. To mitigate these nonlinear and quantization-induced distortions, we propose a learned multi-layer Vector Approximate Message Passing (LMLVAMP) algorithm that incorporates spectral priors with neural network based denoising. Simulation results demonstrate significant performance gains over conventional methods, particularly in high-interference regimes representative of FR3 coexistence scenarios.
Paper Structure (24 sections, 35 equations, 5 figures, 1 table, 1 algorithm)

This paper contains 24 sections, 35 equations, 5 figures, 1 table, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. Classical Nonlinearity Modeling
  4. Classical Interference Suppression
  5. Message Passing Algorithms
  6. Data-Driven Approaches
  7. Motivation for Our Approach
  8. Problem Formulation
  9. Spectral Signal Model
  10. General Nonlinear Model
  11. Soft Saturation Model
  12. Learned ML-VAMP for Nonlinear Filtering
  13. Bayesian ML-VAMP Algorithm
  14. Challenges with ML-VAMP
  15. Learned ML-VAMP Algorithm
  16. ...and 9 more sections

Figures (5)

  • Figure 1: Spectral structure for $L = 2$, where the desired user $\boldsymbol{x}_0$ and interferer $\boldsymbol{x}_1$ occupy disjoint bands $B_0$ and $B_1$. Bars show $\mathrm{Var}[x[k]]$, where $\boldsymbol{x}$ is the DFT of $\boldsymbol{r}$.
  • Figure 2: Block diagram of a nonlinear receiver chain incorporating saturation, receiver noise, and an optional quantizer.
  • Figure 3: Achievable rate versus INR for different SNRs and VAMP iterations. Left columns represent unquantized, and right columns represent quantized observations. Each row shows a specific number of VAMP iterations at 10 dB and 20 dB SNR.
  • Figure : Learned ML-VAMP
  • Figure : Simulation Parameters