CS Paper Radar

Daily listing

2026-11-17 · eess.SP

15 papers first submitted this day (UTC) in eess.SP — Signal Processing. arXiv announces a day's submissions the next evening, so the newest day fills up over ~24 h.

Score = weighted mean ofCitations 20%Upvotes 10%Code stars 10%Venue 15%Code/data 10%Novelty 15%Momentum 15%Substance 5%how it works

AI & Machine Learning 2

Fine-Tuning Gemma 4 Edge Models for MATLAB Signal-Processing Code with SFT and QLoRA on a Laptop iGPU

cs.LG cs.AIcs.SEeess.SP ·2026-11-17 ·Mehmet Altan Toksöz
Abstract · 18 pages, 3 figures, 4 tables

Small language models that run on a laptop are attractive for writing signal-processing code offline, but they make systematic errors on digital signal processing (DSP) conventions. We apply supervised fine-tuning (SFT) to the Gemma 4 edge models E2B and E4B with 66 checked examples covering 11 signal-processing MATLAB subtasks, using low-rank adaptation (LoRA) and quantized LoRA (QLoRA) on the integrated GPU (iGPU) of a laptop, and deploy the 8-bit integer (int8) models with the adapters through OpenVINO. On held-out variants scored by deterministic checkers, the pass rate rises from 18/66 to 37/66 (LoRA) and 37/66 (QLoRA) for E2B and from 32/66 to 47/66 for E4B (QLoRA). Reading all 159 failing programs shows that the adapters remove most errors of convention, such as the sign of a lag, units and array shapes, but few errors in building signals and noise. Applied at run time, the adapter costs 17-19% of the generation speed.

0.38

Pluggable Subband-Decoupled Fusion and Unlabeled Capture Prototype Calibration: Bridging High-SNR Null-Subcarrier Regrowth and Low-SNR Coherent Preambles for Cross-Domain IoT RF Fingerprinting

cs.AI cs.ITeess.ASeess.SP ·2026-11-17 ·Fengchong Yao, Jianbing Li, Qing Liu, Kefeng Song, Haitao Li +2 more
benchmark
Abstract

Zero-overhead physical-layer authentication at edge wireless access points (APs) is vital for securing resource-constrained Internet of Things (IoT) transceivers against credential cloning, yet cross-domain IEEE 802.11a/g radio frequency fingerprint identification (RFFI) faces an acute signal-to-noise ratio (SNR) crossover. Power-amplifier spectral regrowth on unpopulated null subcarriers resists indoor multipath at high SNR but decays quadratically below the thermal noise floor, whereas populated active in-band preambles survive low SNR yet degrade under severe multipath, causing static fusion to lose up to 44.7 percentage points (pp). To our knowledge, we propose the first RFFI framework that dynamically modulates active in-band waveforms and unpopulated null-subcarrier regrowth via a blind physical SNR gate: Pluggable Gated Subband-Decoupled Fusion with Unlabeled Capture-Aware Prototype Calibration (GSDF-CAPC). Avoiding the 20-35 dB subband gradient imbalance of joint training, GSDF pairs any frozen pretrained host backbone with a 46.3K null-subcarrier plug-in encoder (NS-Branch), fusing scale-standardized logits and gate-warped spherical embeddings under a source-calibrated lag-64 thermal-SNR gate with exact low-SNR fallback. Over unlabeled target AP captures, forward-only unbalanced optimal-transport calibration (UOT-CAPC) aligns shifted prototypes under packet-arrival skew and recovers parameter-free balanced Sinkhorn (U-CAPC) at equipartition. Across DataPort-WiFi, WiSig, and ORBIT MobRFFI, zero-retrain GSDF wins all 72/72 paired runs across eight backbones with 0.00 pp low-SNR loss, while edge IB-Host + GSDF (92.6K-93.3K active parameters, 1.82 ms single-core CPU latency) and timing-synchronized GSDF-STO / GSDF-Tri reach 57.50%-65.05% three-benchmark Macro-F1 (89.04%-100.00% over 5-10-packet IoT bursts).

0.27

Language, Vision & Media 1

Personalized Causal Low-Bitrate Coded Speech Restoration with Speaker Embeddings

eess.AS eess.SP ·2026-11-17 ·Eray Özgünay, Christian Rollwage, Kishan Gupta, Andreas Brendel, Xavier Domont +3 more
Abstract · 5 pages, 3 figures, 1 table

Legacy low-bitrate speech codecs such as AMR-NB remain widely used in telephony, where they restrict speech bandwidth and degrade listening quality. Coded speech restoration (CSR) aims to restore the speech bandwidth while improving the overall quality of coded speech, but typically without adapting to different speakers. In this paper, we systematically investigate personalization for restoration of speech coded at low bitrates, using a causal, low-complexity Universal Bandwidth Extension Generative Adversarial Network (UBGAN) model as the basis. We propose two personalization methods based on speaker embeddings, which incur little additional computational cost at inference: (i) conditioning a single CSR model on speaker embeddings, and (ii) cluster-based model selection, where multiple personalized models are trained on clusters of acoustically similar speakers. Experiments are performed on AMR-NB coded speech at 4.75 kbps. Experimental results with disjoint seen- and unseen-speaker test sets show that cluster-based model selection does not consistently outperform the non-personalized UBGAN baseline, whereas embedding conditioning significantly improves ViSQOL and WB-PESQ scores for both seen and unseen speakers.

0.20

Robotics, Control & Signals 12

Least ln-cosh-based adaptive filtering algorithm for graph signal processing

eess.SP ·2026-11-17 ·Chong Zhang, Haiquan Zhao
published
Abstract · Accepted for publication in IEEE Transactions on Aerospace and Electronic Systems

Currently, an increasing number of irregular signals are being defined within graph structures. This paper proposes a Least ln-cosh-based Adaptive Filtering for Graph Signal Processing (GSP LLAF) algorithm for offline/online graph signal estimation under various non-Gaussian noise conditions. Compared to the recently proposed Graph Adaptive Least Mean p-th Power (GSP LMP) algorithm, GSP LLAF eliminates the need for prior knowledge of the noise characteristics. Simultaneously, GSP LLAF addresses the limitation in estimation accuracy encountered by the Sign Algorithm for Graph Signal Processing (GSP Sign) and the Graph Signal Processing algorithm based on the Generalized Maximum Correntropy Criterion (GSP GMCC). Furthermore, the algorithm features low computational complexity and fast execution speed. To achieve a balance between convergence speed and estimation accuracy, a variable step-size variant, the Modified Variable Step-Size Least ln-cosh-based Adaptive Filtering Algorithm for Graph Signal Processing (GSP MVSS-LLAF), is proposed. In estimation experiments involving both static and time-varying graph signals, the proposed GSP LLAF and GSP MVSS-LLAF algorithms demonstrate fast, stable, and robust graph signal estimation performance under various non-Gaussian noise conditions, including Bernoulli-Gaussian (BG) modeled impulse noise, as well as alpha-stable, Cauchy, and Laplace distributions.

0.47

Generation of Sample Complex Wishart Distributed Matrices and Change Detection in Polarimetric SAR Data

eess.SP ·2026-11-17 ·Allan A. Nielsen, Henning Skriver, Knut Conradsen
publishedunusual
Abstract · 7 pages, 15 figures

The complex Wishart distribution is used to describe synthetic aperture radar (SAR) data in the so-called covariance matrix representation. We give a test statistic for detection of differences between two instances in this distribution and an associated probability measure. Generated complex Wishart distributed covariance matrices are used to show that this test statistic and the probability measure in situations with no differences follow the expected distributions.

0.44

Hardware Impairment Parameter Estimation in Fluid Antenna Systems by Inverse Optimization

eess.SP ·2026-11-17 ·Songling Zhang, Zhaohui Yang, Ruopeng Xu, Hao Xu, Xusheng Zhu +2 more
published
Abstract · Accepted by IEEE GLOBECOM 2026

In this paper, the problem of estimating the hardware impairment parameters of fluid antennas (FAs) in a multi-base station (BS) multi-user communication system is studied. The actual hardware coefficients are unknown, whereas the FA positions and transmit beamforming vectors adopted by the BSs are known. We first formulate a forward spectral efficiency (SE) maximization problem and introduce aggregate interference variables whose consistency with the physical inter-cell interference can be maintained through a pricing mechanism. For inverse estimation, the aggregate interference is obtained from receiver-side interference measurements associated with the observed network decisions. We then construct a trust-region linear model of the position-update problem around the observed FA positions. To extend the applicability of the proposed framework to antenna positions obtained from inexact or heuristic algorithms, we develop an inverse optimization model with an explicit duality-gap tolerance, which is solved using sequential quadratic programming (SQP). Simulation results show that the proposed scheme attains a lower estimation error than KKT residual minimization (KKT-RM) at both considered transmit-power levels and improves the SE by up to 22.86% over the Ideal-hardware-assumption scheme.

0.43

Semantic Communication via Analog RF Computing: When the Radio Computes the Meaning

eess.SP cs.AIcs.IT ·2026-11-17 ·Christo Kurisummoottil Thomas
Abstract

Semantic communication (SC) systems transmit continuous latent representations of meaning, but computing the semantic encoder at resource-constrained edge devices incurs substantial energy and memory costs. To address this challenge, we propose an end-to-end SC framework that executes the semantic encoder in the client's RF receiver front end using analog RF computing. A base station (BS) broadcasts encoder weights as RF waveforms, while each edge device reuses its receiver's passive mixer to compute semantic latents at ultra-low energy. These latents are transmitted over the uplink for network-side task inference. We show that analog computing noise across encoder layers and channel noise affecting transmitted latents admit a unified mathematical model. We define a task-posterior semantic distortion metric and derive its second-order decomposition into per-client, per-stage semantic sensitivities. Using this decomposition, we obtain a closed-form semantic-aware water-filling allocation of the total client energy budget across clients, encoder layers, and uplinks, assigning tighter accuracy targets to semantically sensitive stages. This allocation induces a broadcast beamforming problem that accounts for client-specific, per-layer accuracy requirements. We further refine the allocation through stochastic optimization and jointly design it with multi-task noise-aware training of the shared semantic model. MNIST simulations show that, at matched total client energy, the proposed allocation reduces network semantic distortion by 9.2\%-19.2\% compared with uniform precision and 5.1\%-13.1\% compared with client-blind allocation. Noise-aware co-design yields a further 46\% reduction.

0.36

Reference Coordinates, Output Perturbations, and Information Loss in EEG Decoding

eess.SP ·2026-11-17 ·Runni Zhou, Rongrong Fu, Xinghao Yan, Xiaoyang Li
Abstract · 45 pages including supplementary material; 5 figures in the main text. Data and code included as ancillary material

Electroencephalography (EEG) decoders can respond to reference changes that preserve all relative-potential information. We use established measurement-operator theory to distinguish coordinate sensitivity, additional output-channel perturbations, and electrode loss. Exact pseudoinverse alignment supplies a consistency control; measurement-operator canonicalization (MOC) denotes a graph-Tikhonov estimator. The broader evaluation covers three public motor-imagery datasets and four decoders. The primary EEGNet comparison on Cho2017 and PhysioNet uses three paired training seeds, three shared noise realizations, and six front ends, with identical CAR training inputs sharing checkpoints. At 10 dB output perturbation, MOC-minus-raw disagreement was -0.144 and -0.107, while balanced-accuracy differences were 0.044 and 0.050, respectively; these four contrasts passed the retained Holm correction. MOC also reduced disagreement relative to pseudoinversion across all three seeds in both cohorts. However, comparison with fixed graph smoothing after pseudoinversion did not establish an additional MOC benefit. Preserving zero reference rows changed raw disagreement by less than 0.001 in magnitude; inverse-method predictions were unchanged. For canonical front ends, cross-reference disagreement was close to the reference obtained from independent noise within the same operator. Controlled weak-mode reconstruction and equal-rank electrode masks exposed separate stability and information-loss effects. The results locate the complete-reference gains within spatial processing under the specified stress model, while requiring decoding accuracy alongside consistency to assess practical utility.

0.33

Edge-Combining Subharmonic Mixer-First Receiver

eess.SP ·2026-11-17 ·Eric Danson, Jeffrey S. Walling
Abstract · 4 pages, 10 figures, 1 Table

A 1-10 GHz subharmonic mixer-first receiver using edge combining to drive N-path switches from 50% duty-cycle clocks at half the local oscillator (LO) frequency avoids the gain penalty of harmonic mixing. Inverter-based differential baseband amplifiers with back-gate feedback linearization set their gain by process-defined capacitance ratios, using no additional components and consuming no additional power. A 22 nm fully depleted silicon on insulator (FD-SOI) prototype receiver achieves a conversion gain of 12.8-18.1 dB and a noise figure of 5.7-11 dB across the band. It achieves up to 15.4 dBm out-of-band IIP3 while consuming 16.7-22.3 mW.

0.33

Temporal Diffusion Models for Joint Signal-Channel Estimation under Gaussian and $α$-Stable MIMO Channels

eess.SP cs.LG ·2026-11-17 ·Rawad Kansoh, Bahaa Ammoury, Charbel Hannoun, Jihad Fahs, Ibrahim Abou-Faycal
Abstract · 14 pages, 10 figures

Diffusion models have emerged as powerful tools for learning complex data distributions and solving inverse problems. However, most diffusion models rely on the Gaussian distribution, which does not accurately represent heavy-tailed data. In this work, we study diffusion-based reconstruction under Gaussian and symmetric $α$-stable fading channels, while also exploring the benefits of using possible temporal dependence in the channel. We consider two variants: pilot-based channel-only estimation and joint image-channel recovery for MIMO systems. Our approach is based on several standard and score-based diffusion models for both Gaussian and $α$-stable distributions. We refine these models by guiding the reverse diffusion process using the observation likelihood and show how a single trained diffusion model can be reused for the location-scale family of distributions without retraining. In blind MIMO recovery, separate diffusion models are used for the image and channel, and their estimates are combined through likelihood guidance. For each setting, we compare independent and temporal reconstruction, where information from previous observations is used to improve the estimation. We show that, under a time-dependent channel model, such a temporal reconstruction improves channel estimation and the quality of the reconstructed image. Finally, we show that, for the same heavy-tailed channel data, using an $α$-stable diffusion prior provides better channel estimation and image reconstruction than using a Gaussian diffusion prior.

0.33

Structured Design of Layered Sparse Unitary Space-Time Modulation for Integrated Noncoherent and Coherent MIMO Communications

eess.SP math.DG ·2026-11-17 ·Álvaro Pendás-Recondo, Enrique Pendás-Recondo, Rafael González Ayestarán, Jesús Alberto López-Fernández, Ignacio Santamaria
Abstract · 13 pages, 8 figures

This work presents a novel structured constellation design based on unitary space-time modulation (USTM) for integrated noncoherent and coherent multiple-input multiple-output (MIMO) communications. Information is encoded into the sparse space-time matrices through two layers: a noncoherent layer that distinguishes points on the Grassmann manifold (subspaces) and a coherent layer that distinguishes among the Stiefel representatives associated with each of those points (bases for the subspaces). We consider sequential decoding of the two layers to reduce computational complexity, while explicitly identifying the matrix structures that favor this decoding strategy. The proposed constellation builds on a previous structured Grassmannian signaling design, which has the remarkable property of producing space-time matrices with a single nonzero entry per row, while introducing a novel systematic design for the second layer corresponding to the Stiefel representatives. The design of the second layer preserves the sparsity of the first layer and significantly reduces the decoding computational complexity, while ensuring reliable performance under sequential decoding and enabling a simple yet effective bit labeling, as validated by numerical results in terms of bit error rate (BER) and symbol error rate (SER). Numerical results also show graceful performance degradation under channel correlation and, for limited coherence intervals where channel state information at the receiver (CSIR) can be estimated only imperfectly, improved performance compared with classical space-time block coding (STBC) schemes based on QAM and PSK.

0.29

Device-Centric Cooperative Floor-Level Estimation under Sparse Anchor Deployment for Construction-Site Applications

eess.SP ·2026-11-17 ·Jiaxu Liang, Zhuoran Li, Osamu Muta, Haruto Morokuma, Naohiro Ikeda +1 more
Abstract

For construction-site applications in multi-story buildings, accurate floor-level estimation is important for tracking workers, equipment, and Internet of Things (IoT) devices. In buildings under construction, communication infrastructure and power-supply facilities are often limited and the indoor environment changes as construction progresses, making localization systems that rely on dense wireless infrastructure or fingerprint databases difficult to deploy and maintain. Low-power wide-area (LPWA) communication can provide wide-area coverage with a small number of temporarily deployed access points (APs), but under sparse AP deployment, AP-to-terminal observations alone may provide insufficient floor-related information for reliable floor-level estimation. To address this problem, this paper develops a device-centric cooperative floor-level estimation method that exploits inter-terminal observations as relative floor-difference information to complement AP-side information. For practical application with limited measurement data, AP-to-terminal observations are represented by an upper/lower-AP RSS contrast, while inter-terminal observations are statistically modeled according to floor-difference classes. The resulting AP-side and pairwise scores are combined in a weighted floor-label objective and approximately optimized using dual-order sequential iterated conditional modes (ICM). Ray-tracing simulations and LoRa measurement experiments demonstrate the effectiveness of the proposed method under different AP deployments and terminal-selection limits.

0.28

StreamingDPD: A 1.28-GSa/s Instantaneous-Power-Gated Depthwise TCN Accelerator for Real-Time 320-MHz Digital Predistortion

eess.SP cs.AR ·2026-11-17 ·Yizhuo Wu, Xinling Wang, Chang Gao
Abstract

State feedback between successive samples limits the streaming throughput of recurrent neural networks for digital predistortion (DPD). We present StreamingDPD, an accelerator that processes 320-MHz signals at 1.28 GSa/s. Its 207-parameter instantaneous-power-gated depthwise temporal convolutional network (IPG-DWTCN) uses $|x[n]|^2$ to gate feature channels and causal dilated depthwise convolutions to model power amplifier memory. The network maps to a spatially unrolled fixed-point pipeline with shift-based quadratic activations and layer-wise mixed precision, sustaining one complex sample per cycle per lane. Its algorithmic cost is 554 floating-point operations per complex sample, 30.1% fewer than the recurrent baseline with the best linearization performance. In RF playback measurements on three power amplifiers driven by 40- to 320-MHz Wi-Fi 6/7 and 5G NR waveforms, IPG-DWTCN achieves the lowest mean adjacent-channel power ratio (ACPR) and ranks first or ties for first in mean error vector magnitude (EVM) against five neural baselines of comparable size. Four-lane FPGA hardware-in-the-loop tests on a 320-MHz 4096-QAM Wi-Fi 7 waveform achieve -45.82 dBc ACPR and -42.94 dB EVM at 1.28 GSa/s. At the same sample rate, post-layout evaluation of a single-lane 12-nm FinFET ASIC yields a 0.10 mm$^2$ core, a 52-cycle (40.63 ns) latency, and an estimated typical-corner power of 142.21 mW (111.10 pJ/sample). Relative to the prior 22-nm recurrent DPD ASIC, the design provides 5.1 times the throughput at about one-seventh the energy per sample.

0.27

Distributed Blind Source Separation from Compressed Mixtures

eess.SP ·2026-11-17 ·Amir Weiss
benchmarktheory
Abstract

Blind source separation (BSS) is increasingly needed in sensing architectures where mixtures are acquired by multiple remote devices but the inference itself is carried out centrally. In such systems, the sensors often cannot transmit their raw observations since the communication links are rate-limited, energy-limited, or both. The fusion center therefore must recover the latent sources from compressed versions of the mixtures. This setting raises questions that classical BSS models do not address directly: whether source identifiability survives compression, how separation accuracy changes under a specified compression model, and how the separation procedure should be modified to account for the statistical model induced by the compression stage. In this work, we develop a framework for distributed BSS from compressed mixtures that addresses these issues at three levels. First, we formalize identifiability under compression and establish a data-processing viewpoint showing that compression cannot introduce distinguishability that is absent before compression. Second, for mixtures of Gaussian sources with general covariance structures, we derive an idealized Gaussian rate-distortion-based Cramér-Rao lower bound benchmark for the interference-to-source ratio. Third, we develop practical compression-aware separation schemes based on subtractively dithered quantization together with debiasing at the fusion center for both fourth-order-cumulant-based and second-order-statistics-based methods. Simulation experiments demonstrate the resulting tradeoffs between communication rate and separation performance and the gains obtained by compression-aware separation in representative distributed sensing scenarios.

0.26

Robust DOA Estimation for GNSS-like Signals Under Jamming: A Rotation-Enhanced SWN-MUSIC Approach

eess.SP ·2026-11-17 ·Jiyao Jiang, Xiaowei Cui, Gang Liu, Sihao Zhao, Mingquan Lu
Abstract

In jamming environments, accurate direction-of-arrival (DOA) estimation for Global Navigation Satellite System (GNSS)-like signals using antenna arrays is crucial for spoofing detection and subsequent countermeasures. However, estimation accuracy strongly depends on the incident direction of the signal of interest (SOI) under a given jamming and array configuration, with severe errors occurring in certain directions---a phenomenon whose physical mechanisms remain unclear. To address this gap, we identify two types of geometric ill-conditions between the array manifold and the jamming subspace, clarifying that certain directions suffer from inherently unfavorable observation geometry, which cannot be overcome by algorithmic refinement alone. Motivated by this insight, we introduce a paradigm namely rotational diversity, which treats the array attitude as a controllable sensing resource to actively reconfigure the observation geometry, thereby providing complementary perspectives in which the problematic directions can be accurately estimated. To implement this paradigm, we develop a Rotation-Enhanced Spatial Whitening and Normalized Multiple Signal Classification (SWN-MUSIC) framework, which performs independent DOA estimations across multiple rotation angles and selectively integrates reliable estimates. Simulation results demonstrate substantial improvement in accuracy in the directions that were poorly estimated under the fixed-array SWN-MUSIC, while maintaining performance in well-conditioned directions, thereby producing more robust DOA estimates.

0.23