A Deep Generative Model for Synthesizing Labeled Wireless Signals

Yuxiao Li; Keke Hu; Santiago Mazuelas; Yuan Shen

cs.AISep 4, 2026

A Deep Generative Model for Synthesizing Labeled Wireless Signals

Authors: Yuxiao Li, Keke Hu, Santiago Mazuelas, Yuan Shen

Organizations: Basque Center for Applied Mathematics (BCAM), 48009 Bilbao, Spain · IKERBASQUE Foundation for Science, 48009 Bilbao, Spain · College of Semiconductors (College of Integrated Circuits), Changsha 410082, China · Department of Electronic Engineering, Tsinghua University, Beijing 100190, China ·

Abstract

Wireless signals with position-related labels are pivotal for both performance evaluation and model training in the realm of wireless sensing. However, acquiring real-world datasets is often challenged by significant measurement and labeling costs. Traditional methods for synthesizing labeled wireless signals typically rely on environmental models, leading to extensive hyper-parameter tuning and inadequate realism for comprehensive model training purposes. To address these limitations, we introduce a novel deep learning (DL)-based method, namely Inter-Instance Generative Adversarial Networks (IIns-GAN), to generate realistic labeled wireless signals. The generated signals are particularly adaptive to different environment scenarios and well-suited for various model training tasks, including distance estimation and environment identification. We have conducted extensive experiments on public Ultra-Wideband (UWB) datasets to evaluate the realism and utility of the generated signals. The results demonstrate that the signals generated by IIns-GAN mirror the physical characteristics of real-world measurements, and significantly contribute to the improvement of model training in diverse wireless sensing tasks.

Figures & tables

Fig. 1: Signal propagation in different environments: outdoor (LOS), indoor (multipath), and NLOS with wood and metal obstacles. Corresponding received signal waveforms illustrate the impact on signal quality.

Fig. 2: Graphical model for the signal measurement $\bm{x}$ , distance and environmental features $\bm{z}_{\text{d}}$ , $\bm{z}_{\text{e}}$ , as well as distance and environmental labels $d$ and $k$ .

Panel of Figure 3 — Fig. 3: Illustration of adversarial variable $\xi$ and implicit distribution assumption. (a) Graphical model with the global adversarial variable $\xi$ . (b) Variable $\xi$ tells the truthfulness via sampling their empirical distributions from the true and generated observed variable $\bm{x}$ and latent variables $\bm{z}_{\text{d}}$ , $\bm{z}_{\text{e}}$ .

Fig. 4: Network structure of the proposed IIns-GAN. The network parameters are trained on empirical VI objectives in the proposed LVM.

Fig. 5: Illustration of the methodology and usage of two types of realistic signal generations. The upper graph shows the label-based synthesis and the lower graph shows the signal-based translation.

\nabla\mathbb{L}(\mathcal{B};\bm{\phi},\bm{\theta},\bm{\varphi})

Algorithm 1 Variational Learning of IIns-GAN

	$\displaystyle\bm{z}_{\text{d}}^{(n)}\sim$	$\displaystyle\mathcal{N}\big(\bm{z}_{\text{d}};\bm{\mu}(\bm{x}^{(n)},d^{(n)};\bm{\varphi}^{\prime}_{d});\bm{\sigma}^{2}(\bm{x}^{(n)},d^{(n)};\bm{\varphi}^{\prime}_{d})\mathbf{I}\big)$
	$\displaystyle\bm{z}_{\text{e}}^{(n)}\sim$	$\displaystyle\mathcal{N}\big(\bm{z}_{\text{e}};\bm{\mu}(\bm{x}^{(n)},k^{(n)};\bm{\varphi}^{\prime}_{k});\bm{\sigma}^{2}(\bm{x}^{(n)},k^{(n)};\bm{\varphi}^{\prime}_{k})\mathbf{I}\big)$

Algorithm 2 Variational Learning of IIns-GAN

Panel of Figure 8 — Fig. 6: The Signal Waveform of real, synthesized and translated signals.

Panel of Figure 9 — Fig. 7: The PSD comparisons between real and two types of generated signals.

Data Type	Maximum Amplitude (MA)	Rise Time (RT)	Energy	Kurtosis
Synthesized General	0.87	0.44	-0.45	-0.09
Synthesized $d_{1}$	0.88	0.42	-0.46	-0.10
Synthesized $d_{2}$	0.87	0.46	-0.44	-0.07
Synthesized LoS	0.81	0.43	-0.41	-0.01
Synthesized NLoS	0.90	0.44	-0.48	0.03
Synthesized indoor	0.83	0.65	-0.42	0.01

TABLE I: Comparison of real and synthesized signals on four physical features

Data Type	Maximum Amplitude (MA)	Rise Time (RT)	Energy	Kurtosis
Translated General	0.96	0.59	0.87	0.14
Translated $d_{1}$	0.96	0.54	0.86	0.14
Translated $d_{2}$	0.97	0.63	0.87	0.15
Translated LoS	0.97	0.92	0.86	0.06
Translated NLoS	0.97	0.50	0.87	0.08
Translated indoor	0.99	0.81	0.91	-0.29

TABLE II: Comparison of real and translated signals on four physical features

Data Type	MSE (m)	MAE (m)
Real	0.17	0.11
Synthesized ( $\times 1$ )	0.19	0.12
Synthesized ( $\times 10$ )	0.14	0.09
Translated ( $\times 1$ )	0.17	0.11
Translated ( $\times 10$ )	0.14	0.09

TABLE III: Performances of toy CNN on distance estimation, trained on real, synthesized and translated signals.

Data Type	LoS/NLoS Acc	Indoor/Outdoor Acc
Real	0.74	0.67
Synthesized ( $\times 1$ )	0.74	0.64
Synthesized ( $\times 10$ )	0.75	0.74
Translated ( $\times 1$ )	0.74	0.71
Translated ( $\times 10$ )	0.75	0.75

TABLE IV: Performances of toy CNN on environment identification, trained on real, synthesized and translated signals.

Data Type	LoS	NLoS
LoS	0.9	0.7
NLoS	0.7	0.9

TABLE V: Performances before translation.

Data Type	LoS	NLoS
LoS (translated)	0.7	0.9
NLoS (translated)	0.9	0.7

TABLE VI: Performances after translation.

Synthetic Data Generation

Synthetic data generation creates artificial examples, simulations, and datasets to train, test, or protect machine-learning systems.

See all 625 papers

LLM-AUG: Robust Wireless Data Augmentation with In-Context Learning in Large Language Models
Apr 20, 2026Pranshav Gajjar, Manan Tiwari, Sayanta Seth +1Distribution Shift Robustness Synthetic Data Augmentation
NextG Wireless Lab, North Carolina State University, NC
Cross-Modal Generation: From Commodity WiFi to High-Fidelity mmWave and RFID Sensing
Apr 17, 2026Zhixiong Yang, Long Jing, Yao Li +3Synthetic Data Generation Wireless Communications
Tsinghua University Beijing, China
RF-GPT: Teaching AI to See the Wireless World
Feb 16, 2026Hang Zou, Yu Tian, Bohao Wang +4Wireless Communications Language Modeling
Research Institute for Digital Future, Khalifa University, 127788 Abu Dhabi, UAE · College of Information Science and Electronic Engineering, Zhejiang University, 310027, Hangzhou, China · Université de Lorraine, CNRS, CRAN, F-54000 Nancy, France

Apr 20, 2026J/K move · Enter open · S save

LLM-AUG: Robust Wireless Data Augmentation with In-Context Learning in Large Language Models

Pranshav Gajjar, Manan Tiwari, Sayanta Seth, Vijay K. Shah

NextG Wireless Lab, North Carolina State University, NC

Distribution Shift Robustness Synthetic Data Augmentation Wireless Communications Large Language Model-Based Data Augmentation

Open paper AskPDF arXiv

Data scarcity remains a fundamental bottleneck in applying deep learning to wireless communication problems, particularly in scenarios where collecting labeled Radio Frequency (RF) data is expensive, time-consuming, or operationally constrained. This paper proposes LLM-AUG, a data augmentation framework that leverages in-context learning in large language models (LLMs) to generate synthetic training samples directly in a learned embedding space. Unlike conventional generative approaches that require training task-specific models, LLM-AUG performs data generation through structured prompting, enabling rapid adaptation in low-shot regimes. We evaluate LLM-AUG on two representative tasks: modulation classification and interference classification using the RadioML 2016.10A dataset, and the Interference Classification (IC) dataset respectively. Results show that LLM-AUG consistently outperforms traditional augmentation and deep generative baselines across low-shot settings and reaches near oracle performance using only 15% labeled data. LLM-AUG further demonstrates improved robustness under distribution shifts, yielding a 29.4% relative gain over diffusion-based augmentation at a lower SNR value. On the RadioML and IC datasets, LLM-AUG yields a relative gain of 67.6% and 35.7% over the diffusion-based baseline. The t-SNE visualizations further validate that synthetic samples generated by better preserve class structure in the embedding space, leading to more consistent and informative augmentations. These results demonstrate that LLMs can serve as effective and practical data augmenters for wireless machine learning, enabling robust and data-efficient learning in evolving wireless environments.

Abstract

Synthetic Data Generation

Explore similar work

LLM-AUG: Robust Wireless Data Augmentation with In-Context Learning in Large Language Models

Cross-Modal Generation: From Commodity WiFi to High-Fidelity mmWave and RFID Sensing

RF-GPT: Teaching AI to See the Wireless World