FlowSep-RF: Rectified Flow Matching for Single-Channel Radio-Frequency Signal Separation

Mohamed Kamel; Tamer Nadeem

eess.SPOct 14, 2026

FlowSep-RF: Rectified Flow Matching for Single-Channel Radio-Frequency Signal Separation

Authors: Mohamed Kamel, Tamer Nadeem

Organizations: Department of Computer Science, Virginia Commonwealth University, Richmond, VA, USA

Abstract

Single-channel radio-frequency (RF) signal separation is important for interference mitigation when a signal of interest overlaps with a co-channel interferer in time and frequency. Classical matched filtering and LMMSE receivers are limited under structured non-Gaussian interference, while diffusion and score-based methods often require iterative sampling or optimization. We propose FlowSep-RF, a rectified flow matching framework that learns a clean-endpoint transport from the received mixture to the signal of interest. The model combines time-domain IQ samples with frequency-domain features and recovers the separated waveform using one endpoint velocity evaluation. We evaluate FlowSep-RF on QPSK mixtures with OFDM-BPSK, OFDM-QPSK, and RFChallenge CommSignal2 interference against classical, discriminative, and generative baselines. FlowSep-RF achieves the lowest average BER and MSE among evaluated baselines across all three settings with one-step inference.

Figures & tables

Figure 1: FlowSep-RF overview. Training forms a clean-endpoint path $x_{t}=(1-t)y+ts$ from the mixture $y$ to the clean SOI $s$ and learns $v^{\star}=s-y$ using time- and frequency-domain features conditioned on RF variables. Inference uses estimated RF conditions and one endpoint update, $\hat{s}=y+v_{\theta}(y,y,0,\hat{\gamma}_{\mathrm{SIR}},\hat{\kappa})$ , followed by demodulation for BER evaluation.

Panel of Figure 2 — Figure 2: BER and MSE versus SIR for FlowSep-RF and baseline methods on the three interference settings. The top row reports BER after demodulation, and the bottom row reports waveform MSE. Lower is better.

Method	OFDM-BPSK		OFDM-QPSK		CommSignal2
	BER	MSE	BER	MSE	BER	MSE
MF	0.186716	–	0.189842	–	0.196826	–
LMMSE + MF	0.186716	0.904915	0.189842	0.904157	0.196826	0.904333
Reverse Diffusion	0.015707	0.066848	0.037840	0.166426	0.387684	0.511738
$\alpha$ -RGS	0.001711	0.007473	0.008855	0.021780	0.066684	0.139131
DirectComm-RF	0.013484	28.138059	0.006269	28.037159	0.027385	7.222556

Table 1: Main comparison on QPSK SOI separation. BER/MSE are averaged over the test SIR grid. Lower is better; dash indicates unavailable or non-applicable.

Variant	OFDM-BPSK		OFDM-QPSK		CommSignal2
	BER	MSE	BER	MSE	BER	MSE
Full FlowSep-RF	0.000645	0.001470	0.003776	0.006231	0.023066	0.034845
w/o rectified flow (direct reg.)	0.000920	0.002150	0.005120	0.009410	0.031200	0.045100
w/o symbol-proxy loss	0.000782	0.001615	0.004120	0.006892	0.025140	0.038102
w/o spectral loss	0.000715	0.001850	0.004050	0.007420	0.024800	0.040150
w/o SIR weighting	0.000810	0.001740	0.004310	0.007100	0.026100	0.039800

Table 2: Ablation study. BER and MSE are averaged over the test SIR grid. Lower is better.

Wireless Communications

Wireless communications studies how information is sent over radio links and how networks manage signals, spectrum, and resources reliably.

See all 391 papers

The Radio-Frequency Transformer for Signal Separation
Mar 10, 2026Egor Lifar, Semyon Savkin, Rachana Madhukara +3Finite Scalar Quantization Transformer
Massachusetts Institute of Technology (MIT).
SURF: Separation via Unsupervised Remixing Flow
Jun 3, 2026Henry Li, Robin Scheibler, Efthymios Tzinis +3Unsupervised Learning Flow Matching
Google · Google DeepMind
Restoration Flow Matching-Based Channel Refinement and Equalization Correction for MIMO Semantic Communications
Jul 26, 2026Wenkai Liu, Nan Ma, Jianqiao Chen +3Flow Matching Semantic Communication
State Key Laboratory of Networking and Switching Technology, Beijing University of Posts and Telecommunications, Beijing 100876, China · ZGC Institute of Ubiquitous-X Innovation and Applications, Beijing 100876, China · Cooperative Medianet Innovation Center (CMIC) and the Department of Electronic Engineering, Shanghai Jiao Tong University, Shanghai 200240, China

Mar 10, 2026J/K move · Enter open · S save

The Radio-Frequency Transformer for Signal Separation

Egor Lifar, Semyon Savkin, Rachana Madhukara, Tejas Jayashankar+2

Massachusetts Institute of Technology (MIT).

Finite Scalar Quantization Transformer

Open paper AskPDF arXiv

We study a problem of signal separation: estimating a signal of interest (SOI) contaminated by an unknown non-Gaussian background/interference. Given the training data consisting of examples of SOI and interference, we show how to build a fully data-driven signal separator. To that end we learn a good discrete tokenizer for SOI and then train an end-to-end transformer on a cross-entropy loss. Training with a cross-entropy shows substantial improvements over the conventional mean-squared error (MSE). Our tokenizer is a modification of Google's SoundStream, which incorporates additional transformer layers and switches from VQVAE to finite-scalar quantization (FSQ). Across real and synthetic mixtures from the MIT RF Challenge dataset, our method achieves competitive performance, including a 122x reduction in bit-error rate (BER) over prior state-of-the-art techniques for separating a QPSK signal from 5G interference. The learned representation adapts to the interference type without side information and shows zero-shot generalization to unseen mixtures at inference time, underscoring its potential beyond RF. Although we instantiate our approach on radio-frequency mixtures, we expect the same architecture to apply to gravitational-wave data (e.g., LIGO strain) and other scientific sensing problems that require data-driven modeling of background and noise.

Abstract

Wireless Communications

Explore similar work

The Radio-Frequency Transformer for Signal Separation

SURF: Separation via Unsupervised Remixing Flow

Restoration Flow Matching-Based Channel Refinement and Equalization Correction for MIMO Semantic Communications