Heuristics beat lattices for OFDM pilot patterns
A-optimal sensor selection solved by convex or greedy methods with local swaps yields lower estimation error than fixed grids on realistic,
eess.SP
Signal Processing
Theory, algorithms, performance analysis and applications of signal and data analysis, including physical modeling, processing, detection and parameter estimation, learning, mining, retrieval, and information extraction. The term "signal" includes speech, audio, sonar, radar, geophysical, physiological, (bio-) medical, image, video, and multimodal natural and man-made signals, including communication signals and data. Topics of interest include: statistical signal processing, spectral estimation and system identification; filter design, adaptive filtering / stochastic learning; (compressive) sampling, sensing, and transform-domain methods including fast algorithms; signal processing for machine learning and machine learning for signal processing applications; in-network and graph signal processing; convex and nonconvex optimization methods for signal processing applications; radar, sonar, and sensor array beamforming and direction finding; communications signal processing; low power, multi-core and system-on-chip signal processing; sensing, communication, analysis and optimization for cyber-physical systems such as power grids and the Internet of Things.
sort pith recommended most recent
ChannelLM predicts wireless channels in new places with 4.23 dB less error
It rebuilds the physical scene from cameras and sensors then uses a large model to forecast signal behavior for real-time 6G use.
Synthetic radar dataset provides 6000 pulse trains for deinterleaving
Nearly 3 billion pulses from up to 110 emitters create a public benchmark for electronic warfare signal processing.
· “The Turing Synthetic Radar Dataset: A dataset for pulse deinterleaving”
Crossing counts estimate ARMA poles via autocorrelation
Only event tallies from the series need storage to recover poles for assessing control loop behavior.
· “A New Approach for ARMA Pole Estimation Using Higher-Order Crossings”
Differentiable GNSS solver learns credible covariances
Supervising the predictive distribution with scoring rules cuts mean horizontal error from 13.77 m to 11.68 m on harsh-urban test data.
· “CredibleDFGO: Differentiable Factor Graph Optimization with Credibility Supervision”
Power series models near-field wavefronts as polynomials
This recasts channel estimation over antenna arrays as a multidimensional polynomial phase problem, improving the accuracy-complexity trade
· “Near-field channel estimation via wavefront parameterization”
The paper describes an automatic method using differentiable beamforming to correct…
A differentiable beamforming framework optimizes aberration correction using angular coherence, improving transcranial ULM and fUS imaging…
Prior knowledge cuts spurious change points in time series
A time-varying penalty derived from prior location information slots into the fast PELT algorithm and improves accuracy without losing exact
· “Pi-Change: A Prior-Informed Multiple Change Point Detection Algorithm”
This paper reviews recent advancements in mm-wave oscillators below 100 GHz and…
A literature review summarizing design approaches, performance metrics, challenges, and emerging techniques for mm-wave and sub-THz/THz…
· “Recent Advances in mm-Wave and Sub-THz/THz Oscillators for FutureG Technologies”
Two-layer analog network matches digital beamforming in multi-user MISO
Lossless reciprocal two-layer architecture delivers identical sum-rate to fully digital methods.
· “Two-Layer Microwave Linear Analog Computer (MiLAC)-aided Multi-user MISO Networks”
Physics embedding discovers optimal hidden parameters
Training solely on reconstruction quality lets a neural estimator find unobservable values that compensate for model errors in wireless and
· “Discovery of unobservable parameters via physical embedding”
Separable method betters Doppler estimates in passive radar
Delay accuracy matches the joint approach while Doppler gains and computation plus communication costs fall for slow targets.
Peak power constraints degrade AirComp federated learning
Clipping signals to stay within amplifier limits introduces distortions that reduce accuracy, particularly in multi-carrier systems.
· “On Signal Peak Power Constraint of Over-the-Air Federated Learning”
Causal networks recover global sections when Laplacian spectra allow
Sheaf theory coordinates imperfect causal models across agents without shared graphs or joint data, with diffusion converging to consistent,
· “Networks of Causal Abstractions: A Sheaf-theoretic Framework”
Compact model cuts 5G channel estimation time by 45%
Matches leading accuracy under low SNR while using eight times fewer parameters and running in half the time.
· “HELENA: High-Efficiency Learning-based channel Estimation using dual Neural Attention”
Switching off idle radio heads cuts C-RAN power by 20 percent
A joint RRH-association, fronthaul, and power-allocation loop beats max-SINR and saves over 20% power in dense networks.
· “Power-Efficient Resource Allocation in Massive MIMO Aided Cloud RANs”
Resolution after dictionary learning cannot beat min(s
Even with deployment data in hand, calibration noise sets the floor: relative precision scales as N times the sixth power of s.
· “Physical-Support Confidence Sets for Highly Coherent Dictionaries”
E-detectors exactly capture ARL-safe alarm times
Any alarm rule with average run length at least b is a threshold crossing of an e-detector; strong rules also bound every adaptive horizon.
PHALAR model lifts stem retrieval accuracy by up to 70 percent with half the parameters
Contrastive framework adds pitch and phase equivariance via spectral pooling and complex head, trains seven times faster, and matches human
· “PHALAR: Phasors for Learned Musical Audio Representations”
Transformer doubles radar resolution on real-world scenes
1D shifted-window attention and CFAR losses yield sharp range-Doppler maps at low compute cost.
· “Super-Resolution of Range-Doppler Maps: A Case Study with Chirp-Sequence Radar and Transformer”
Splits signals into components without a known count
IDSD learns component shapes from synthetic data, then separates tidal waves from tsunamis and cleans noisy pulse signals.
· “IDSD: Iterative Deep-Learning-Based Signal Decomposition”
Adaptive clustering lifts wireless delivery by up to 9.8 points
Grouping peers by reliability and coding within and across clusters cuts latency 10-23% and raises retention up to 30%.
Fusion gates spot lost sensors
In an eight-strategy ECG-PPG fusion comparison, gates reweight sharply at dropout but barely track graded corruption.
A WiFi array becomes an AR camera for invisible radio
Live channel data reveals signal direction, delay, and polarization overlaid on the camera view.
· “Visualizing Wireless Propagation and Polarization in Augmented Reality with ESPARGOS”
Receiver learns PA distortion from one OFDM symbol
Maximum-likelihood DCT estimator gives near-ideal BER with 2,000–20,000 training samples, down to −10 dB SNR.
· “Efficient DCT-Based Estimation and Compensation of Nonlinear Channels for OFDM Systems”
Cough-based TB models don't transfer across datasets
External AUCs often drop below 0.6 even when internal scores hit 0.75, pointing to device bias.
Learned aggregation matches FedAvg and Ditto on non-IID FL
Server and client agents jointly learn aggregation weights and personalization without pretraining.
Bed vibrations track blood pressure within AAMI error limits
Triaxial bodyseismography plus physics-constrained deep learning gives SBP/DBP/MAP MAEs of 6.0, 4.2, 5.1 mmHg
Inverse-design pipeline hits target bands on all four test antennas
Surrogate-guided latent-space search yields connected pixel antennas that resonate inside a chosen 22–30 GHz band.
· “Machine Learning Assisted Inverse Design of Pixelated mmWave Patch Antennas”
Phase-blind detector tracks the optimal maximum-likelihood receiver
A rule using only the RIS channel's mean matches the optimal detector; more elements postpone the error floor.
· “RIS-Assisted Radar-Communication Coexistence: Detection Analysis with Channel Uncertainties”
Five RF paths shift phase a full 360 degrees
Resumming weighted, delayed copies replaces quadrature networks, keeping magnitude flat for SATCOM beam steering
Full-band ISAC waveform images at 30 cm while carrying MIMO data
Multipath-aware equalization recovers streams despite strong reflections; 60-GHz tests confirm ranging and the zero-correlation region.
At -50 dB, learned RF fusion still detects objects
Three-panel RF imaging fused by a U-Net keeps semantic labels and boxes usable where classical reconstruction has turned to noise.
· “Semantic Reconstruction and 3-D Detection via Learned Multi-Pair Fusion in RF Imaging”
Noise-free boundary data locate hidden nonlinear anomalies
Monotonicity plus the p-Laplace limit reconstructs outer support or convex hull, even without uniform material contrast.
Curved pinching antennas get a closed-form optimal placement rule
Where straight waveguides fail, a curved ceiling layout can help; the rule says exactly where to place the antenna.
· “Curved Waveguide-Enabled Pinching-Antenna System (C-PAS): Communication Performance Analysis”
Centralizing LoRa processing buys up to 4.2 dB of sensitivity
Joint baseband combining across remote radio heads improves reception for static and moving IoT transmitters.
· “Centralized RAN for Future Low-Power Wide-Area Networks: A LoRa Case Study”
Curved smart surfaces lift indoor and outdoor link success by 75–125%
Facade curvature widens which users a single surface can serve, but denser cities shrink the gain.
· “Simultaneous Indoor and Outdoor Coverage with Conformal Intelligent Omni-Surface”
Channel-token policy tops 92 percent access success
A Transformer that reads each channel as a token beats greedy and PPO baselines, with the edge widening as primary-user load rises.
· “Channel-Token Attention for Reliable Dynamic Spectrum Access under Bursty Primary-User Traffic”
AirMoE aggregates MoE expert outputs over the air in one shot
A threshold power rule and rank-based placement keep end-to-end accuracy high when device channels differ.
· “AirMoE: Realizing Over-the-Air Distributed Mixture-of-Experts Inference at the Wireless Edge”
Two-sided AI works with legacy 5G: 4x throughput at low SNR
Plus a model table for channel adaptation and gradient-free fine-tuning that cuts communication 15,000x.
· “Rethinking the Foundations of Two-Sided AI Models for 6G”
This paper proposes IMNet, a learning-based receiver that removes two types of subcarrier…
IMNet and IMNet-LR make SEFDM ISAC practical by jointly suppressing SEFDM-induced and Doppler-induced intercarrier interference, achieving…
PAINT is a new network that converts 7-band Landsat satellite images into 172-band…
PAINT is an interpretable deep-unfolding network for Landsat-to-AVIRIS dual super-resolution that improves PSNR by about 2 dB over the best…
· “Interpretable Landsat-to-Hyperspectral Dual Super-Resolution Without Large Matrix Inversion”
A new framework treats leveraged ETFs on correlated stocks as one coupled feedback system
Market stability of leveraged products should be judged by a matrix measuring coupled feedback across correlated assets, not by each…
Drill-down cuts geospatial anomaly scans by 99.7%
Start coarse, zoom only where anomalies appear: 4,800 cells inspected instead of 2 million.
Baseband circuits, not RF filters, key to FR3 receivers
A design review shows selectivity, linearity, and noise fixes can be translated from baseband to 7-24 GHz.
· “Interference-Tolerant Mixer-First Receivers for FR3: Design Principles and Tradeoffs”
Six-fold speedup: ISAC beamforming needs only channel statistics
A Courant-Fischer bound shows when the positive-semidefinite projection is unnecessary, cutting runtime up to six times.
· “ISAC Beamforming Design Based on a Matrix Nearness Formulation With Improved Efficiency”
ISAC beam design drops from hours to 1.3 seconds
A matrix-nearness trick plus accelerated dual updates gives cheaper beampatterns and per-antenna PAPR caps.
· “Waveform Design for Simultaneous MIMO Radar Sensing and Multi-User Communication”
Metasurface antenna streams live 720p video over 60 GHz
A software-defined radio link with off-the-shelf front-ends carries about 3 Mbit/s over one meter
Low-band pilots predict sub-THz delays to within 10 ns
Deep learning maps sub-10 GHz channels to sub-THz delays, making switches seamless without two-way sync.
Unrolled network sharpens 3D radio holograms with 3,090 parameters
Learned sparse priors and a low-rank Born correction beat back-projection on measured body scenes.
· “EM Informed Holographic Imaging via Unrolled Deep Networks”
20% of labels beat 100% in wearable emotion recognition
Self-supervised graph masking lifts arousal by 4.3% and valence by 7.8% on K-EmoPhone.
Using coherence structure cuts the DOA bound
New proof shows the deterministic Cramer-Rao bound shrinks only when multipath coherence is modeled in detail.
Closed-form updates match near-optimal beamforming at 1% of the runtime
Per-AP and per-SIM-layer updates replace general-purpose solvers in weighted sum-rate beamforming.
Spiking AI matches CNN on sonar at a third of the energy
A fully spiking detector uses 3.3–6.2x less theoretical energy across three forward-looking sonar datasets and resists speckle noise.
· “Spiking Neural Networks for Energy-Efficient Object Detection in Forward-Looking Sonar Imagery”
Matching vibrato makes two instruments sound like one
A suppress-then-transfer chain aligns per-harmonic AM, FM, and residual modulation until unison sources fuse.
· “Vibrato Matching for Modulation Control and Blending in Sound Mixtures”
Chirp-sensing moving antenna cuts indoor mmWave outage to 0.9%
The pinching antenna senses moving humans, repositions itself, and keeps immersive links at 2850-3000 Mbps.
· “Sensing-Assisted Anti-Blockage Pinching-Antenna Systems For Indoor Immersive Communications”
Physics-aware loss maps radio channels without GPS or timestamps
It switches distances for LoS and NLoS, beating seven baselines in ray-traced tests.
· “PACC: Propagation-Aware Channel Charting with Physics-Guided Metric Learning”
Masking private pixels before wireless transmission halves leakage
A server-side vision-language model rebuilds the removed regions while eavesdroppers stay at a 48% perceptual deficit.
· “Privacy Preserving Semantic Communications in Wireless Edge Networks with Vision Language Models”
Payload samples track near-field beams at 0.97 gain
A Thompson-sampling tracker reuses feedback symbols, avoiding dedicated beam sweeps even under sharp-turn mobility.
· “Adaptive Payload-Aided Near-Field Beam Tracking via Thompson Sampling”
Force energy timetables measurement info in diffusion inversion
Expected squared smoothed-likelihood force equals entropy dissipation, giving one exact budget for when measurements act.
· “Posterior Information Dynamics of Diffusion Models for Linear Inverse Problems”
First survey maps RIS-based cognitive radio networks
Six research directions, six numerical case studies, and a 6G roadmap in one reference.
COPD voice signal survives age matching
Age-free acoustic models keep 0.717 AUC on cohorts where raw age and gender are at chance.
Pattern shortcut understates slot-array sidelobes by 1–3 dB
Full-wave check of a 16×16 WR-90 array shows edge coupling defeats element-pattern times array-factor.
A one-vector beamformer beats MIMO radar beampattern design
A new algorithm shapes transmit beams with deeper nulls and lower sidelobes in milliseconds, not minutes.
· “Transmit Beamformer Design for Beampattern Synthesis in Phased-Array Radar Systems”
HAPs plus UAV swarms could replace ground base stations for 6G
HAP-plus-UAV cell-free architecture promises autonomous 6G coverage and spectrum sharing with terrestrial networks.
· “HAP-Centric Flying Ad-Hoc Networks With Cell-Free Non-Terrestrial Connectivity”
Congruence decomposition beats 5G PCI baselines by up to 53%
Splitting cell IDs into prime-power blocks, a GNN solver cuts interference and runs 2x faster on real networks.
· “Congruence Decomposition with Neural Block Solvers for Large-Scale PCI Assignment”
Obstacle-aware planning keeps receivers' line of sight to the target
Steering receivers to preserve line of sight avoids the tracking failures that static receivers suffer.
· “Obstacle-Aware Online Receiver Planning in Multistatic Ranging”
Wireless AI works better small if physics guides it
A survey argues that shared radio backbones can beat big models by embedding signal physics into learning.
Drones with movable smart surfaces boost secure ISAC secrecy
Six-dimensional pose control beats fixed and rotation-only reflecting surfaces in line-of-sight networks.
· “Secrecy Rate Maximization for UAV-Mounted Six-Dimensional Movable IRS-Assisted ISAC Systems”
OTFS keeps BER within 2 dB of ideal despite RF impairments
A single low-PAPR pilot handles timing, CFO, phase noise, and transmitter IQ imbalance together.
· “Synchronization and Channel Estimation of OTFS with RF Impairments”
MTSFM optimizer matches wideband MIMO beampattern at 50% bandwidth
A closed-form cross-spectral density lets the optimizer shape constant-modulus MIMO beams at fractional bandwidth 0.5.
· “Wideband MIMO Beampattern Synthesis using Adaptive Frequency Modulated Waveforms”
Modeling the multipath background rescues weak sonar tracks
Including the tracked background in the likelihood confirms weak sonar targets that thresholded detectors miss.
0.49 mW chip decodes auditory attention from EEG
A 0.076 mm² CNN plus classifier runs in 7.34 ms, fitting the 8 ms real-time EEG budget.
· “A Resource-Efficient CNN-Based EEG Auditory Attention Decoding ASIC”
Zero-crossing LLRs match exact soft decisions in RIN links
A simple linear LLR rule avoids up to 7.2x BER penalty and 2.41 dB OMA loss.
· “Low-complexity Soft-decision LLR Calculations for Next-generation IM-DD Systems with RIN”
Moving antennas can send extra bits through their own Doppler shift
A CSI-free detector reads three velocity-index bits about 10 dB earlier than pilot-charged position indexing.
Longer GNSS integration cuts error
Tighter correlators and 10 ms integration cut error, but switching modes can drop the signal.
· “Some Practical Issues of the Tracking Process in GNSS Receivers”
Four capabilities cover wireless resilience
Robustness, adaptability, survivability, recoverability: the hierarchy that ties them to mechanisms and trade-offs.
Gradient optimization beats random search 72-0 in cavity design
Matched on time and dielectric budget, gradient-guided FDTD put more spectral energy in the target band in every trial.
Feature transmission in FL is twice as private as gradient transmission
Send only encoder features and clients get the same privacy with half the noise, under one uplink budget
Co-designing feeds and surface amplitudes suppresses radar sidelobes
An alternating SCA method brings a reconfigurable holographic surface near fully-digital benchmarks in OFDM-ISAC.
· “Holographic Beamforming for Range-Doppler Sidelobe Suppression in OFDM-ISAC”
