Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Learning to Separate RF Signals Under Uncertainty: Detect-Then-Separate vs. Unified Joint Models

Ariel Rodrigez, Alejandro Lancho, Amir Weiss

TL;DR

The paper investigates how to separate a signal of interest from interference in a single-channel RF setting when the interference type is uncertain. It analyzes a detect-then-separate (DTS) approach, proven to be asymptotically MMSE-optimal in a Gaussian mixture model under a temporal-diversity condition, and compares it to a unified joint model (UJM) that jointly detects and separates using a single deep network. Through analysis and UNet-based experiments on synthetic and RF Challenge data, the authors show that a capacity-matched UJM can match oracle-aided DTS performance across diverse SIR levels, interference types, and modulation orders, highlighting scalability and robustness to type uncertainty. These results suggest unified models as a practical alternative to DTS, enabling scalable, robust RF signal separation in complex interference environments.

Abstract

The increasingly crowded radio frequency (RF) spectrum forces communication signals to coexist, creating heterogeneous interferers whose structure often departs from Gaussian models. Recovering the interference-contaminated signal of interest in such settings is a central challenge, especially in single-channel RF processing. Existing data-driven methods often assume that the interference type is known, yielding ensembles of specialized models that scale poorly with the number of interferers. We show that detect-then-separate (DTS) strategies admit an analytical justification: within a Gaussian mixture framework, a plug-in maximum a posteriori detector followed by type-conditioned optimal estimation achieves asymptotic minimum mean-square error optimality under a mild temporal-diversity condition. This makes DTS a principled benchmark, but its reliance on multiple type-specific models limits scalability. Motivated by this, we propose a unified joint model (UJM), in which a single deep neural architecture learns to jointly detect and separate when applied directly to the received signal. Using tailored UNet architectures for baseband (complex-valued) RF signals, we compare DTS and UJM on synthetic and recorded interference types, showing that a capacity-matched UJM can match oracle-aided DTS performance across diverse signal-to-interference-and-noise ratios, interference types, and constellation orders, including mismatched training and testing type-uncertainty proportions. These findings highlight UJM as a scalable and practical alternative to DTS, while opening new directions for unified separation under broader regimes.

Learning to Separate RF Signals Under Uncertainty: Detect-Then-Separate vs. Unified Joint Models

TL;DR

The paper investigates how to separate a signal of interest from interference in a single-channel RF setting when the interference type is uncertain. It analyzes a detect-then-separate (DTS) approach, proven to be asymptotically MMSE-optimal in a Gaussian mixture model under a temporal-diversity condition, and compares it to a unified joint model (UJM) that jointly detects and separates using a single deep network. Through analysis and UNet-based experiments on synthetic and RF Challenge data, the authors show that a capacity-matched UJM can match oracle-aided DTS performance across diverse SIR levels, interference types, and modulation orders, highlighting scalability and robustness to type uncertainty. These results suggest unified models as a practical alternative to DTS, enabling scalable, robust RF signal separation in complex interference environments.

Abstract

The increasingly crowded radio frequency (RF) spectrum forces communication signals to coexist, creating heterogeneous interferers whose structure often departs from Gaussian models. Recovering the interference-contaminated signal of interest in such settings is a central challenge, especially in single-channel RF processing. Existing data-driven methods often assume that the interference type is known, yielding ensembles of specialized models that scale poorly with the number of interferers. We show that detect-then-separate (DTS) strategies admit an analytical justification: within a Gaussian mixture framework, a plug-in maximum a posteriori detector followed by type-conditioned optimal estimation achieves asymptotic minimum mean-square error optimality under a mild temporal-diversity condition. This makes DTS a principled benchmark, but its reliance on multiple type-specific models limits scalability. Motivated by this, we propose a unified joint model (UJM), in which a single deep neural architecture learns to jointly detect and separate when applied directly to the received signal. Using tailored UNet architectures for baseband (complex-valued) RF signals, we compare DTS and UJM on synthetic and recorded interference types, showing that a capacity-matched UJM can match oracle-aided DTS performance across diverse signal-to-interference-and-noise ratios, interference types, and constellation orders, including mismatched training and testing type-uncertainty proportions. These findings highlight UJM as a scalable and practical alternative to DTS, while opening new directions for unified separation under broader regimes.
Paper Structure (11 sections, 2 theorems, 18 equations, 6 figures, 1 table)

This paper contains 11 sections, 2 theorems, 18 equations, 6 figures, 1 table.

Table of Contents

  1. Introduction
  2. Problem Formulation
  3. SCSS Data-Driven Approaches with Interference Uncertainty
  4. The Detect-Then-Separate Approach
  5. The Unified Joint Model Approach
  6. Analytical Analysis for the Gaussian Mixture Signal Model
  7. Simulation Results
  8. Simulation Setup
  9. Numerical Results
  10. Concluding Remarks
  11. Lemma 1 Proof Outline

Key Result

Lemma 1

proofs Under the TDC, for any finite $\alpha>0$ independent of $N$,

Figures (6)

  • Figure 1: Block diagrams of the two DNN architectures for signal separation under interference uncertainty.
  • Figure 2: QPSK SOI with mixtures containing CS3 ($p=1/2$): BER vs. SINR for DTS, 5L-UJM and UJM.
  • Figure 3: QPSK SOI with mixtures containing CS5G1 ($p=1/2$): BER vs. SINR for DTS, 5L-UJM and UJM.
  • Figure 4: QPSK SOI with CS3 interference only ($p=1$): BER and MSE vs. SINR for DTS, 5L-UJM and UJM.
  • Figure 5: $M$PSK SOI with EMIS1 or CS2 interference ($p=1$ or $0$, respectively): BER vs. SINR for DTS, 5L-UJM and UJM.
  • ...and 1 more figures

Theorems & Definitions (2)

  • Lemma 1
  • Theorem 1