arxiv: 2604.10183 · v1 · submitted 2026-04-11 · 💻 cs.DC · cs.LG

Recognition: unknown

RF-LEGO: Modularized Signal Processing-Deep Learning Co-Design for RF Sensing via Deep Unrolling

Luca Jiang-Tao Yu , Chenshu Wu

Authors on Pith no claims yet

Pith reviewed 2026-05-10 15:38 UTC · model grok-4.3

classification 💻 cs.DC cs.LG

keywords deep unrollingRF sensingsignal processingdeep learningmodular designwireless sensinginterpretability

0 comments

The pith

RF-LEGO converts signal processing algorithms into modular deep learning blocks via deep unrolling for reusable and interpretable RF sensing.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper establishes a co-design approach that takes established signal processing routines and unfolds them into trainable neural modules while keeping their core mathematical structure. This produces components for frequency analysis, angle estimation, and detection that can be combined or reused across tasks instead of requiring separate end-to-end models. A sympathetic reader would value the result because it promises both higher accuracy than hand-tuned processing and clearer internal operation than typical deep networks. The authors demonstrate the gains on real measurements from Wi-Fi, millimeter-wave, UWB, and 6G sensing scenarios, both when modules run alone and when inserted into larger pipelines.

Core claim

RF-LEGO applies deep unrolling to three RF sensing primitives so that fixed signal-processing operators become layered structures whose parameters are learned from data; the resulting modules remain physics-aligned and composable, delivering higher performance than standalone signal processing or task-specific deep networks on real-world traces while preserving interpretability through retained operator structure.

What carries the argument

Deep unrolling, the process of mapping iterative signal-processing steps into a fixed-depth neural network whose layers mirror the original operators but contain learnable parameters.

If this is right

The modules can be cascaded or swapped to build new sensing pipelines without starting from random weights.
Performance advantages appear both in isolated use and when the blocks are inserted into larger downstream applications.
Interpretability follows directly from the preserved processing flow rather than post-hoc explanation techniques.
The same unrolling recipe applies across Wi-Fi, millimeter-wave, UWB, and emerging 6G sensing systems.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

If the modules remain composable, pre-trained blocks could be mixed to address new sensing problems with far less labeled data than training full models from scratch.
Hardware mapping of the fixed operators could yield efficient accelerators that retain the learned parameters without full neural-network overhead.
The approach might generalize to other signal domains where iterative algorithms already exist, such as radar imaging or acoustic processing.

Load-bearing premise

That replacing fixed parameters in signal-processing routines with learned ones inside an unrolled structure will preserve the original mathematical behavior and physical meaning while producing reliable gains across different wireless environments and tasks.

What would settle it

Running the three unrolled modules on a fresh real-world RF dataset for frequency transform, angle estimation, or detection and finding that accuracy falls below or matches conventional non-learned signal processing baselines.

Figures

Figures reproduced from arXiv: 2604.10183 by Chenshu Wu, Luca Jiang-Tao Yu.

**Figure 1.** Figure 1: Core principles of RF-LEGO: Modularity, Cascadability, and Interpretability. RF-LEGO bridges the gap between classical SP and DL for RF sensing via deep unrolling. and semantically meaningful intermediate outputs in standard signal domains. Recently, deep unrolling techniques have emerged as a promising approach to SP-DL co-design by unrolling classical algorithms in a learnable framework. Deep unrolling… view at source ↗

**Figure 2.** Figure 2: Deep Unrolling. S: Signal Processing Block; O: Unrolled Trainable Operator; I: Unrolled Trainable Iterative Block. 2 A PRIMER ON DEEP UNROLLING Most deep learning models are purely data-driven, and their learned structures are difficult to interpret. End-to-end networks learn task-specific mappings (e.g., regression or classification) entirely through backpropagation over the highdimensional parameters … view at source ↗

**Figure 3.** Figure 3: RF-LEGO FT. (a) Bluestein’s Algorithm implementation of Fourier transformation via convolution, where 𝒃 represents the chirp signal and 𝒛 ∗ denotes its conjugate. (b) RF-LEGO FT replaces the fixed convolution with a learnable convolutional layer. Deep Unrolling. The convolutional form in Eqn. (2) reveals a practical handle: the discrete Fourier Transformation can be executed as a single convolution with a … view at source ↗

**Figure 4.** Figure 4: RF-LEGO Beamformer. (a) The classical LASSO solver using ADMM. (b) The unrolled architecture with learnable parameters. Here, 𝒙 (𝒕) represents the estimated sparse angular spectrum, 𝒛 (𝒕) is the auxiliary variable, 𝒗 (𝒕) is the dual variable, and 𝒈 (𝒕) denotes the learned gate that dynamically balances the update between historical and current estimates. complex Gaussian noise 𝒏 ∼ CN (0, 𝜎2 𝑰). The recei… view at source ↗

**Figure 5.** Figure 5: RF-LEGO Detector. (a) Classical CFAR uses a fixed sliding window. (b) RF-LEGO unrolls the logic into a state space model, where 𝒓 is the input signal vector, 𝒛 represents the learned state vector, and 𝒔ˆ is the detection vector derived from the state. At each step, a specific sample, denoted 𝒓𝑛, is designated as the cell under test, while its surrounding samples, excluding a guard region, form a local nei… view at source ↗

**Figure 6.** Figure 6: Experimental scenarios. (a-f) Range experiments; (g-h) Doppler experiments; (g,i) Angle experiments. radar [42], which operates at a center frequency of 7.29 GHz and a frame rate of 100 Hz. • Wi-Fi. Wi-Fi signals are acquired using a commercial router and an Intel AX200 NIC. The transmitter antennas send packets to two receiver antennas, enabling the extraction of channel state information (CSI). To accur… view at source ↗

**Figure 7.** Figure 7: The results of RF-LEGO Range FT of mmWave and RF-LEGO Doppler FT of mmWave, UWB, and Wi-Fi. Scenario g Scenario i [PITH_FULL_IMAGE:figures/full_fig_p009_7.png] view at source ↗

**Figure 11.** Figure 11: Microbenchmarks. (a) Performance on public datasets. (b) Impact of multiple targets. (c) Impact of data fine-tuning [PITH_FULL_IMAGE:figures/full_fig_p011_11.png] view at source ↗

**Figure 13.** Figure 13: Module behavior analysis. (a) RF-LEGO FT learns a non-uniform convolution kernel beyond Bluestein’s Algorithm FT. (b) RF-LEGO Beamformer learns adaptive gate schedules across unrolled iterations. (c) RF-LEGO Detector learns a structured state matrix 𝑨, unlike memoryless CFAR. 13(c) compares the learned state matrix 𝑨 with the memoryless CFAR baseline. While CFAR corresponds to a null state transition, RF… view at source ↗

**Figure 15.** Figure 15: Case study scenarios. (a-b) Trajectory tracking; (c-e) Vital sign monitoring for infant simulator and human adults [PITH_FULL_IMAGE:figures/full_fig_p012_15.png] view at source ↗

**Figure 18.** Figure 18: CDF of breathing MAE. Infant Simulator Human Adult [PITH_FULL_IMAGE:figures/full_fig_p012_18.png] view at source ↗

**Figure 19.** Figure 19: Performance for an infant simulator and the human adult. 0 20 40 60 80 100 Time [s] 10 20 30 Respiration Rate [BPM] Ground Truth Signal Processing RF-LEGO [PITH_FULL_IMAGE:figures/full_fig_p012_19.png] view at source ↗

read the original abstract

Wireless sensing, traditionally relying on signal processing (SP) techniques, has recently shifted toward data-driven deep learning (DL) to achieve performance breakthroughs. However, existing deep wireless sensing models are typically end-to-end and task-specific, lacking reusability and interpretability. We propose RF-LEGO, a modular co-design framework that transforms interpretable SP algorithms into trainable, physics-grounded DL modules through deep unrolling. By replacing hand-tuned parameters with learnable ones while preserving core processing structures and mathematical operators, RF-LEGO ensures modularity, cascadability, and structure-aligned interpretability. Specifically, we introduce three deep-unrolled modules for critical RF sensing tasks: frequency transform, spatial angle estimation, and signal detection. Extensive experiments using real-world data for Wi-Fi, millimeter-wave, UWB, and 6G sensing demonstrate that RF-LEGO significantly outperforms existing SP and DL baselines, both standalone and when integrated into multiple downstream tasks. RF-LEGO pioneers a novel SP-DL co-design paradigm for wireless sensing via deep unrolling, shedding light on efficient and interpretable deep wireless sensing solutions. Our code is available at https://github.com/aiot-lab/RF-LEGO.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

RF-LEGO offers modular deep-unrolled modules that blend signal processing and deep learning for RF sensing tasks.

read the letter

The key takeaway is that RF-LEGO introduces a way to modularize RF sensing by deep-unrolling signal processing algorithms into deep learning modules. Specifically, they created three modules for frequency transform, spatial angle estimation, and signal detection that can be used standalone or cascaded for other tasks. This approach is new in how it emphasizes reusability and interpretability in the SP-DL co-design for wireless sensing. Most DL models for this are end-to-end and task-specific, but here the modules keep the mathematical structures and operators from traditional SP while replacing fixed parameters with learnable ones. That should make them more adaptable and easier to understand than pure neural nets. The paper shows solid results on real-world datasets covering Wi-Fi, millimeter-wave, UWB, and 6G sensing scenarios. It reports better performance than both traditional SP techniques and existing DL models, and the modules integrate well into multiple downstream applications. Having the code publicly available on GitHub is helpful for reproduction and further work. There are a couple of areas that could use more attention. The abstract highlights outperformance but doesn't include specific numbers or detailed comparisons, so the full paper needs to demonstrate that the improvements are consistent and not due to particular choices in the experimental setup. Also, while deep unrolling is meant to retain core properties, it would be good to see explicit checks on whether interpretability holds after training or if the modules behave as expected in varied conditions. Overall, this paper targets people working on practical wireless sensing systems, particularly those interested in combining the strengths of signal processing and machine learning for better efficiency and reliability. Readers who deal with RF data in IoT or 6G contexts will likely find the modular design useful for their own implementations. Given the novelty of the framework and the empirical support, it deserves to go through peer review rather than being rejected at the desk. My recommendation is to send it for peer review. The ideas are grounded and the experiments span multiple technologies, which makes it worth a closer look by experts in the field.

Referee Report

2 major / 3 minor

Summary. The paper proposes RF-LEGO, a modular SP-DL co-design framework that applies deep unrolling to convert interpretable signal processing algorithms into trainable, physics-grounded DL modules for RF sensing. It introduces three specific modules (frequency transform, spatial angle estimation, signal detection) that replace hand-tuned parameters with learnable ones while retaining core operators and structures, enabling modularity and cascadability. Experiments on real-world Wi-Fi, mmWave, UWB, and 6G datasets claim significant outperformance over standalone SP and DL baselines, both as standalone modules and when integrated into downstream tasks, with code released for reproducibility.

Significance. If the empirical gains are robustly documented, the work could meaningfully advance interpretable and reusable deep models for wireless sensing by providing a principled bridge between traditional SP and DL. The emphasis on modularity, preservation of mathematical operators, and public code release are strengths that could support broader adoption and extension across sensing tasks.

major comments (2)

Abstract and §4 (Experiments): the claim of 'significantly outperforms existing SP and DL baselines' is load-bearing for the central contribution, yet the abstract provides no quantitative metrics, specific baseline implementations, or error bars; the experiments section must include full tables with exact numbers, statistical significance tests, and controls for dataset selection to rule out post-hoc bias.
§3 (Module Design): the assertion that core mathematical properties are preserved after unrolling (e.g., in the frequency transform and angle estimation modules) lacks an explicit verification step or invariant check; without this, the interpretability claim rests on structural similarity rather than demonstrated equivalence.

minor comments (3)

Notation consistency: ensure that learnable parameters introduced in the unrolled modules are uniformly denoted (e.g., avoid mixing θ and W across equations).
Figure clarity: the block diagrams for cascaded modules should include explicit input/output dimensions and data flow arrows for each RF sensing task.
Missing reference: add citation to the original deep-unrolling literature (e.g., the foundational works on unrolling iterative algorithms) when describing the transformation process.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments, which help clarify the presentation of our contributions. We address each major point below and will incorporate revisions to strengthen the manuscript.

read point-by-point responses

Referee: Abstract and §4 (Experiments): the claim of 'significantly outperforms existing SP and DL baselines' is load-bearing for the central contribution, yet the abstract provides no quantitative metrics, specific baseline implementations, or error bars; the experiments section must include full tables with exact numbers, statistical significance tests, and controls for dataset selection to rule out post-hoc bias.

Authors: We agree that the abstract would benefit from explicit quantitative support for the performance claims. We will revise the abstract to include key metrics (e.g., accuracy or error reductions with error bars) from the main experiments. In §4, we will expand the tables to report exact numerical results for all baselines, add statistical significance tests (such as paired t-tests with p-values), and include a dedicated paragraph detailing dataset selection criteria, preprocessing steps, and controls to mitigate selection bias. Baseline implementations are already specified in the text and code release, but we will add cross-references for clarity. revision: yes
Referee: §3 (Module Design): the assertion that core mathematical properties are preserved after unrolling (e.g., in the frequency transform and angle estimation modules) lacks an explicit verification step or invariant check; without this, the interpretability claim rests on structural similarity rather than demonstrated equivalence.

Authors: We acknowledge that an explicit verification step would make the preservation of mathematical properties more rigorous. Although the deep-unrolling construction retains the original operators and structures by design (ensuring properties such as linearity in the frequency transform or orthogonality constraints in angle estimation), we will add a new subsection in §3 (or an appendix) that provides both a brief mathematical argument for invariance and empirical checks, such as comparing module outputs on synthetic inputs before and after training to confirm equivalence within numerical tolerance. revision: yes

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The paper's derivation applies the established technique of deep unrolling to convert existing SP algorithms (frequency transform, angle estimation, detection) into modular DL modules by replacing hand-tuned parameters with learnable ones while retaining core operators and structures. All performance claims rest on empirical comparisons against SP and DL baselines using real-world Wi-Fi/mmWave/UWB/6G datasets, with code released for reproducibility. No load-bearing step reduces a prediction to a fitted input by construction, invokes self-citations for uniqueness theorems, or renames known results; the framework is externally testable and self-contained against benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract provides no explicit free parameters, axioms, or invented entities; the approach relies on transforming existing SP algorithms without introducing new postulated components.

pith-pipeline@v0.9.0 · 5523 in / 1002 out tokens · 26080 ms · 2026-05-10T15:38:08.233033+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

81 extracted references · 14 canonical work pages · 1 internal anchor

[1]

Ahmed Alkhateeb, Gouranga Charan, Tawfik Osman, Andrew Hredzak, Joao Morais, Umut Demirhan, and Nikhil Srinivas. 2023. DeepSense 6G: A large-scale real-world multi-modal sensing and com- munication dataset.IEEE Communications Magazine61, 9 (2023), 122– 128

2023
[2]

2000.Digital processing of signals: theory and prac- tice

Maurice Bellanger. 2000.Digital processing of signals: theory and prac- tice. John Wiley & Sons

2000
[3]

Leo Bluestein. 2003. A linear filtering approach to the computation of discrete Fourier transform.IEEE Transactions on Audio and Electroa- coustics18, 4 (2003), 451–455

2003
[4]

Mohammud J Bocus, Wenda Li, Shelly Vishwakarma, Roget Kou, Chong Tang, Karl Woodbridge, Ian Craddock, Ryan McConville, Raul Santos-Rodriguez, Kevin Chetty, et al. 2022. OPERAnet, a multimodal activity recognition dataset acquired from radio frequency and vision- based sensors.Scientific data9, 1 (2022), 474

2022
[5]

Mohammud J Bocus and Robert Piechocki. 2022. A comprehensive ultra-wideband dataset for non-cooperative contextual sensing.Scien- tific Data9, 1 (2022), 650

2022
[6]

Stephen Boyd, Neal Parikh, Eric Chu, Borja Peleato, Jonathan Eckstein, et al. 2011. Distributed optimization and statistical learning via the alternating direction method of multipliers.Foundations and Trends® in Machine learning3, 1 (2011), 1–122

2011
[7]

Jack Capon. 1969. High-resolution frequency-wavenumber spectrum analysis.Proc. IEEE57, 8 (1969), 1408–1418

1969
[8]

Ricky TQ Chen, Yulia Rubanova, Jesse Bettencourt, and David K Duve- naud. 2018. Neural ordinary differential equations.Advances in neural information processing systems31 (2018)

2018
[9]

Yunjin Chen and Thomas Pock. 2016. Trainable nonlinear reaction diffusion: A flexible framework for fast and effective image restoration. IEEE transactions on pattern analysis and machine intelligence39, 6 (2016), 1256–1272

2016
[10]

Zhe Chen, Tianyue Zheng, Chao Cai, and Jun Luo. 2021. MoVi-Fi: Motion-robust vital signs waveform recovery via deep interpreted RF sensing. InProceedings of the 27th annual international conference on mobile computing and networking. 392–405

2021
[11]

Guoxuan Chi, Zheng Yang, Chenshu Wu, Jingao Xu, Yuchong Gao, Yunhao Liu, and Tony Xiao Han. 2024. RF-diffusion: Radio signal generation via time-frequency diffusion. InProceedings of the 30th Annual International Conference on Mobile Computing and Networking. 77–92

2024
[12]

Junyoung Chung, Caglar Gulcehre, KyungHyun Cho, and Yoshua Bengio. 2014. Empirical evaluation of gated recurrent neural networks on sequence modeling.arXiv preprint arXiv:1412.3555(2014)

work page internal anchor Pith review arXiv 2014
[13]

James W Cooley and John W Tukey. 1965. An algorithm for the machine calculation of complex Fourier series.Mathematics of compu- tation19, 90 (1965), 297–301

1965
[14]

Shuya Ding, Zhe Chen, Tianyue Zheng, and Jun Luo. 2020. RF-net: A unified meta-learning framework for RF-enabled one-shot human activity recognition. InProceedings of the 18th Conference on Embedded Networked Sensor Systems. 517–530

2020
[15]

Tzvi Diskin, Yiftach Beer, Uri Okun, and Ami Wiesel. 2024. CFARNet: Deep Learning for Target Detection with constant false alarm rate. Signal Processing223 (2024), 109543

2024
[16]

Laura Dodds, Tara Boroushaki, Kaichen Zhou, and Fadel Adib. 2025. Non-Line-of-Sight 3D Object Reconstruction via mmWave Surface Normal Estimation. (2025)

2025
[17]

Xiangyu Gao, Youchen Luo, Guanbin Xing, Sumit Roy, and Hui Liu
[18]

https://doi.org/10.21227/xm40-jx59

Raw ADC Data of 77GHz MMWave radar for Automotive Object Detection. https://doi.org/10.21227/xm40-jx59

work page doi:10.21227/xm40-jx59
[19]

Xiangyu Gao, Guanbin Xing, Sumit Roy, and Hui Liu. 2021. RAMP- CNN: A Novel Neural Network for Enhanced Automotive Radar Object Recognition.IEEE Sensors Journal21, 4 (2021), 5119–5132. https: //doi.org/10.1109/JSEN.2020.3036047

work page doi:10.1109/jsen.2020.3036047 2021
[20]

Karol Gregor and Yann LeCun. 2010. Learning fast approximations of sparse coding. InProceedings of the 27th international conference on international conference on machine learning. 399–406

2010
[21]

Albert Gu, Isys Johnson, Karan Goel, Khaled Saab, Tri Dao, Atri Rudra, and Christopher Ré. 2021. Combining recurrent, convolutional, and continuous-time models with linear state space layers.Advances in neural information processing systems34 (2021), 572–585

2021
[22]

Arjun Gupta, Dashiell Kosaka, Edwin Pan, Jingning Tang, Ruihao Yao, and Sanjay Patel. 2019. OpenRadar: A Toolkit for Prototyping mmWave Radar Applications.arXiv preprint arXiv:1912.12395(2019)

work page arXiv 2019
[23]

James D Hamilton. 1994. State-space models.Handbook of econometrics 4 (1994), 3039–3080

1994
[24]

Gregers Hansen

V. Gregers Hansen. 1973. Constant false alarm rate processing in search radars. InIEE Conference on Radar, Present and Future. 325–332

1973
[25]

Weiying Hou and Chenshu Wu. 2024. Rfboost: Understanding and boosting deep wifi sensing via physical data augmentation.Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies 8, 2 (2024), 1–26

2024
[26]

Norden E Huang, Zheng Shen, Steven R Long, Manli C Wu, Hsing H Shih, Quanan Zheng, Nai-Chyuan Yen, Chi Chao Tung, and Henry H Liu. 1998. The empirical mode decomposition and the Hilbert spectrum for nonlinear and non-stationary time series analysis.Proceedings of the Royal Society of London. Series A: mathematical, physical and engineering sciences454, 19...

1998
[27]

Daisuke Ito, Satoshi Takabe, and Tadashi Wadayama. 2019. Trainable ISTA for sparse signal recovery.IEEE Transactions on Signal Processing 67, 12 (2019), 3113–3125

2019
[28]

Sijie Ji, Xinzhe Zheng, and Chenshu Wu. 2024. Hargpt: Are llms zero- shot human activity recognizers?. In2024 IEEE International Workshop on Foundation Models for Cyber-Physical Systems & Internet of Things (FMSys). IEEE, 38–43

2024
[29]

Haowen Lai, Gaoxiang Luo, Yifei Liu, and Mingmin Zhao. 2024. En- abling visual recognition at radio frequency. InProceedings of the 30th Annual International Conference on Mobile Computing and Networking. 388–403

2024
[30]

Shuheng Li, Ranak Roy Chowdhury, Jingbo Shang, Rajesh K Gupta, and Dezhi Hong. 2021. Units: Short-time fourier inspired neural networks for sensory time series classification. InProceedings of the 19th ACM conference on embedded networked sensor systems. 234–247

2021
[31]

Yuelong Li, Mohammad Tofighi, Junyi Geng, Vishal Monga, and Yon- ina C Eldar. 2020. Efficient and interpretable deep blind image de- blurring via algorithm unrolling.IEEE Transactions on Computational Imaging6 (2020), 666–681

2020
[32]

Yadong Li, Dongheng Zhang, Jinbo Chen, Jinwei Wan, Dong Zhang, Yang Hu, Qibin Sun, and Yan Chen. 2022. DI-Gesture: Domain- Independent and Real-Time Gesture Recognition with Millimeter- Wave Signals. InGLOBECOM 2022 - 2022 IEEE Global Communications Conference. 5007–5012. https://doi.org/10.1109/GLOBECOM48099.20 22.10001175

work page doi:10.1109/globecom48099.20 2022
[33]

Yadong Li, Dongheng Zhang, Jinbo Chen, Jinwei Wan, Dong Zhang, Yang Hu, Qibin Sun, and Yan Chen. 2023. Towards Domain- Independent and Real-Time Gesture Recognition Using mmWave Sig- nal.IEEE Transactions on Mobile Computing22, 12 (2023), 7355–7369. https://doi.org/10.1109/TMC.2022.3207570

work page doi:10.1109/tmc.2022.3207570 2023
[34]

Qiushi Liang, Yeyue Cai, Jianhua Mo, and Meixia Tao. 2025. CFARNet: Learning-Based High-Resolution Multi-Target Detection for Rainbow Beam Radar.arXiv preprint arXiv:2505.10150(2025). RF-LEGO MobiCom ’26, October 26–30, 2026, Austin, TX, USA

work page arXiv 2025
[35]

Chia-Hung Lin, Yu-Chien Lin, Yue Bai, Wei-Ho Chung, Ta-Sung Lee, and Heikki Huttunen. 2019. DL-CFAR: A novel CFAR target detection method based on deep learning. In2019 IEEE 90th Vehicular Technology Conference (VTC2019-Fall). IEEE, 1–6

2019
[36]

2010.Nonlinear digital filters: analysis and applica- tions

Wing-Kuen Ling. 2010.Nonlinear digital filters: analysis and applica- tions. Academic Press

2010
[37]

Zichao Long, Yiping Lu, Xianzhong Ma, and Bin Dong. 2018. Pde- net: Learning pdes from data. InInternational conference on machine learning. PMLR, 3208–3216

2018
[38]

Sheng Lyu and Chenshu Wu. 2024. ASE: Practical Acoustic Speed Estimation Beyond Doppler via Sound Diffusion Field.arXiv preprint arXiv:2412.20142(2024)

work page arXiv 2024
[39]

Dmitry Malioutov, Müjdat Cetin, and Alan S Willsky. 2005. A sparse signal reconstruction perspective for source localization with sensor arrays.IEEE transactions on signal processing53, 8 (2005), 3010–3022

2005
[40]

Julian P Merkofer, Guy Revach, Nir Shlezinger, Tirza Routtenberg, and Ruud JG Van Sloun. 2023. DA-MUSIC: Data-driven DoA estimation via deep augmented MUSIC algorithm.IEEE Transactions on Vehicular Technology73, 2 (2023), 2771–2785

2023
[41]

Vishal Monga, Yuelong Li, and Yonina C Eldar. 2021. Algorithm un- rolling: Interpretable, efficient deep learning for signal and image processing.IEEE Signal Processing Magazine38, 2 (2021), 18–44

2021
[42]

(DBA OptiTrack) NaturalPoint

Inc. (DBA OptiTrack) NaturalPoint. 2025. OptiTrack — Motion Capture Systems. https://www.optitrack.com/

2025
[43]

Novelda AS. 2018. XeThru X4 Radar User Guide. https://github.c om/novelda/Legacy-Documentation/blob/master/Application- Notes/XTAN-13_XeThruX4RadarUserGuide_rev_a.pdf. Application Note XTAN-13, Rev. A

2018
[44]

Carl M O’Brien. 2016. Statistical learning with sparsity: the lasso and generalizations. (2016)

2016
[45]

Muhammed Zahid Ozturk, Chenshu Wu, Beibei Wang, Min Wu, and KJ Ray Liu. 2023. Radio SES: mmwave-based audioradio speech en- hancement and separation system.IEEE/ACM Transactions on Audio, Speech, and Language Processing31 (2023), 1333–1347

2023
[46]

PyTorch Core Team. 2024. torch.linalg.eig — PyTorch Documentation. https://pytorch.org/docs/stable/generated/torch.linalg.eig.html. Accessed: 2025-04-15

2024
[47]

Guy Revach, Nir Shlezinger, Xiaoyong Ni, Adria Lopez Escoriza, Ruud JG Van Sloun, and Yonina C Eldar. 2022. KalmanNet: Neural network aided Kalman filtering for partially known dynamics.IEEE Transactions on Signal Processing70 (2022), 1532–1547

2022
[48]

Hermann Rohling. 2007. Radar CFAR thresholding in clutter and multiple target situations.IEEE transactions on aerospace and electronic systems4 (2007), 608–621

2007
[49]

Ralph Schmidt. 1986. Multiple emitter location and signal parameter estimation.IEEE transactions on antennas and propagation34, 3 (1986), 276–280

1986
[50]

Fei Shang, Panlong Yang, Dawei Yan, Sijia Zhang, and Xiang-Yang Li
[51]

ACM Interact

LiquImager: Fine-grained Liquid Identification and Container Imaging System with COTS WiFi Devices.Proc. ACM Interact. Mob. Wearable Ubiquitous Technol.8, 1, Article 15 (March 2024), 29 pages. https://doi.org/10.1145/3643509

work page doi:10.1145/3643509 2024
[52]

Sharad Singhal and Lance Wu. 1988. Training multilayer perceptrons with the extended Kalman algorithm.Advances in neural information processing systems1 (1988)

1988
[53]

Oren Solomon, Regev Cohen, Yi Zhang, Yi Yang, Qiong He, Jianwen Luo, Ruud JG van Sloun, and Yonina C Eldar. 2019. Deep unfolded robust PCA with application to clutter suppression in ultrasound.IEEE transactions on medical imaging39, 4 (2019), 1051–1063

2019
[54]

Long Tan, Weijie Yuan, Xiaoqi Zhang, Kecheng Zhang, Zhongjie Li, and Yonghui Li. 2024. DNN-Based Radar Target Detection With OTFS. IEEE Transactions on Vehicular Technology(2024)

2024
[55]

Texas Instruments. 2025. DCA1000EVM Evaluation Module. https: //www.ti.com/tool/DCA1000EVM. Accessed: June 2, 2025

2025
[56]

Texas Instruments. 2025. IWR1843BOOST Evaluation Module. https: //www.ti.com/tool/IWR1843BOOST. Accessed: June 2, 2025

2025
[57]

2025.TI mmWave Labs - Vital Signs Measurement (version 1.2)

Texas Instruments. 2025.TI mmWave Labs - Vital Signs Measurement (version 1.2). Texas Instruments. Accessed: Sep. 2, 2025. Available at: https://e2e.ti.com/cfs-file/__key/communityserver-discussions- components-files/1023/vitalSigns_5F00_lab_5F00_user_5F00_guide _5F00_v1.2UPDATE.pdf

2025
[58]

HL Van Trees et al. 2002. Optimum array processing

2002
[59]

GV Truck. 1977. Range resolution of targets using automatic detectors. NASA STI/Recon Technical Report N78 (1977), 21348

1977
[60]

2004.Detection, estimation, and modulation theory, part I: detection, estimation, and linear modulation theory

Harry L Van Trees. 2004.Detection, estimation, and modulation theory, part I: detection, estimation, and linear modulation theory. John Wiley & Sons

2004
[61]

Fengyu Wang, Feng Zhang, Chenshu Wu, Beibei Wang, and KJ Ray Liu. 2020. ViMo: Multiperson vital sign monitoring using commodity millimeter-wave radio.IEEE Internet of Things Journal8, 3 (2020), 1294–1307

2020
[62]

M Weiss. 2007. Analysis of some modified cell-averaging CFAR pro- cessors in multiple-target situations.IEEE Trans. Aerospace Electron. Systems1 (2007), 102–114

2007
[63]

2022.High-dimensional data analysis with low-dimensional models: Principles, computation, and applications

John Wright and Yi Ma. 2022.High-dimensional data analysis with low-dimensional models: Principles, computation, and applications. Cam- bridge University Press

2022
[64]

Kailun Wu, Yiwen Guo, Ziang Li, and Changshui Zhang. 2020. Sparse coding with gated learned ISTA. InInternational conference on learning representations

2020
[65]

Dawei Yan, Panlong Yang, Fei Shang, Feiyu Han, Yubo Yan, and Xiang- Yang Li. 2025. Pushing the Limits of WiFi-Based Gait Recognition Towards Non-Gait Human Behaviors .IEEE Transactions on Mobile Computing01 (Feb. 2025), 1–17. https://doi.org/10.1109/TMC.2025.3 540863

work page doi:10.1109/tmc.2025.3 2025
[66]

Zheng Yang, Yi Zhang, Kun Qian, and Chenshu Wu. 2023. {SLNet}: A Spectrogram Learning Neural Network for Deep Wireless Sensing. In 20th USENIX Symposium on Networked Systems Design and Implemen- tation (NSDI 23). 1221–1236

2023
[67]

Shuochao Yao, Ailing Piao, Wenjun Jiang, Yiran Zhao, Huajie Shao, Shengzhong Liu, Dongxin Liu, Jinyang Li, Tianshi Wang, Shaohan Hu, et al. 2019. Stfnets: Learning sensing signals from the time-frequency perspective with short-time fourier neural networks. InThe World Wide Web Conference. 2192–2202

2019
[68]

Luca Jiang-Tao Yu, Running Zhao, Sijie Ji, Edith CH Ngai, and Chen- shu Wu. 2025. USpeech: Ultrasound-Enhanced Speech with Minimal Human Effort via Cross-Modal Synthesis.Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies9, 2 (2025), 1–31

2025
[69]

Feng Zhang, Chenshu Wu, Beibei Wang, Min Wu, Daniel Bugos, Hang- fang Zhang, and KJ Ray Liu. 2019. SMARS: Sleep monitoring via ambient radio signals.IEEE Transactions on Mobile Computing20, 1 (2019), 217–231

2019
[70]

Haoyun Zhang, Jianghong Han, Xueqian Wang, Gang Li, and Xiao- Ping Zhang. 2025. Parameter Convergence Detector Based on VAMP Deep Unfolding: A Novel Radar Constant False Alarm Rate Detection Algorithm.arXiv preprint arXiv:2504.09912(2025)

work page arXiv 2025
[71]

Haopeng Zhang, Yili Ren, Haohan Yuan, Jingzhe Zhang, and Yitong Shen. 2025. Wi-chat: Large language model powered wi-fi sensing. arXiv preprint arXiv:2502.12421(2025)

work page arXiv 2025
[72]

Xie Zhang, Yina Wang, and Chenshu Wu. 2025. Unlocking Inter- pretability for RF Sensing: A Complex-Valued White-Box Transformer. arXiv preprint arXiv:2507.21799(2025). MobiCom ’26, October 26–30, 2026, Austin, TX, USA Yu and Wu

work page arXiv 2025
[73]

Yuwei Zhang, Kumar Ayush, Siyuan Qiao, A Ali Heydari, Girish Narayanswamy, Maxwell A Xu, Ahmed A Metwally, Shawn Xu, Jake Garrison, Xuhai Xu, et al. 2025. SensorLM: Learning the Language of Wearable Sensors.arXiv preprint arXiv:2506.09108(2025)

work page arXiv 2025
[74]

Yi Zhang, Weiying Hou, Zheng Yang, and Chenshu Wu. 2023. {VeCare}: Statistical acoustic sensing for automotive{In-Cabin} mon- itoring. In20th USENIX Symposium on Networked Systems Design and Implementation (NSDI 23). 1185–1200

2023
[75]

Mingmin Zhao, Yonglong Tian, Hang Zhao, Mohammad Abu Alsheikh, Tianhong Li, Rumen Hristov, Zachary Kabelac, Dina Katabi, and Anto- nio Torralba. 2018. RF-based 3D skeletons. InProceedings of the 2018 Conference of the ACM Special Interest Group on Data Communication. 267–281

2018
[76]

Peijun Zhao, Chris Xiaoxuan Lu, Bing Wang, Niki Trigoni, and Andrew Markham. 2023. Cubelearn: End-to-end learning for human motion recognition from raw mmwave radar signals.IEEE Internet of Things Journal10, 12 (2023), 10236–10249

2023
[77]

Running Zhao, Jiangtao Yu, Hang Zhao, and Edith CH Ngai. 2023. Radio2text: Streaming speech recognition using mmwave radio signals. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies7, 3 (2023), 1–28

2023
[78]

Running Zhao, Luca Jiang-Tao Yu, Tingle Li, Zhihan Jiang, Chenwei Zhang, Chenshu Wu, Hang Zhao, and Edith CH Ngai. 2025. SPACE: Speaker Adaptation for Acoustic Eavesdropping using mmWave Radio Signals.IEEE Transactions on Mobile Computing(2025)

2025
[79]

Xiaopeng Zhao, Zhenlin An, Qingrui Pan, and Lei Yang. 2023. Nerf2: Neural radio-frequency radiance fields. InProceedings of the 29th An- nual International Conference on Mobile Computing and Networking. 1–15

2023
[80]

Tianyue Zheng, Zhe Chen, Shujie Zhang, Chao Cai, and Jun Luo. 2021. MoRe-Fi: Motion-robust and fine-grained respiration monitoring via deep-learning UWB radar. InProceedings of the 19th ACM conference on embedded networked sensor systems. 111–124

2021

Showing first 80 references.