Signal Processing
See recent articles
Showing new listings for Friday, 20 November 2026
- [1] arXiv:2611.25054 [pdf, html, other]
-
Title: Adaptive Filtering With Reduced Computational Complexity Using SOPOT ArithmeticJournal-ref: IEEE Transactions on Circuits and Systems I: Regular Papers (Volume: 69, Issue: 2, February 2022)Subjects: Signal Processing (eess.SP)
Implementing finite impulse response (FIR) adaptive filters by employing the sums of signed-powers-of-two (SOPOT) arithmetic may lead to simpler hardware and consequently reduced power consumption. In this paper, one evaluates the effects of SOPOT arithmetic on the adaptive filter's recursion algorithms. The filters' coefficients and algorithms' underlying variables are fully operated using SOPOT arithmetic in the whole iterative process. More specifically, one evaluates convergence rate, numerical stability, and accuracy since using few signed-powers-of-two (SPT) terms propagates numerical errors during the adaptive cycle that may impair the algorithm behavior. The SOPOT approximations are obtained through the technique known as Matching Pursuits with Generalized Bit-Plane (MPGBP) algorithm, with notable cost-performance trade-off and low computational complexity. Results are provided for the Least-Mean-Squares (LMS), the Normalized Least-Mean-Squares (NLMS) and the Recursive-Least-Squares (RLS) algorithms, considering adaptive filters employed for system identification and change detection.
- [2] arXiv:2611.25262 [pdf, html, other]
-
Title: AE-DISC: Automorphism Ensemble Direct-Invocation Successive Cancellation for Low-Latency DecodingSubjects: Signal Processing (eess.SP)
Emerging applications in hyper-reliable low-latency communication (HRLLC) demand channel decoders that deliver deterministic performance under stringent latency and reliability constraints. While Polar codes are well suited for this scenario due to their strong error-correction capability, conventional successive cancellation (SC) decoding exhibits inherent serial dependencies that lead to high decoding latency. This paper presents the automorphism ensemble direct-invocation successive cancellation (AE-DISC) decoder, a low-latency framework that enables the concurrent computation of the left and right decoding subtrees in the SC process, yielding substantial speed-ups. By exploiting permutation automorphisms, AE-DISC decoding further improves error-correction performance. Simulation results demonstrate that AE-DISC attains up to 1.4 dB gain over the 5G SC baseline and surpasses layered min-sum (LMS) low-density parity-check (LDPC) decoders at a frame error rate (FER) of $10^{-6}$, while maintaining fixed algorithmic complexity. Both theoretical analysis and application-specific integrated circuit (ASIC) synthesis results further show up to 2 times decoding speed-up relative to fast simplified successive cancellation (FSSC).
- [3] arXiv:2611.25304 [pdf, html, other]
-
Title: Conditional-Hierarchical Open-Set Deep Learning for Robust Radio Signals IntelligenceSubjects: Signal Processing (eess.SP)
This paper proposes an open-set radio-frequency (RF) deep learning (DL) framework that jointly performs conditionally hierarchical multi-level radio signal classification and out-of-distribution (OOD) detection. To exploit image-based classification models, the raw radio samples are transformed into time-frequency representations using the short time Fourier transform (STFT) to pre-process the radio signals. A lightweight novel architecture using neural network (CNN), that applies con- ditional hierarchical multi-task classification heads, is introduced, and compared to two baselines including: (1) The conventional convolutional neural network (CNN) and (2) Swin-T model with multitask classification heads, where the model accuracy and complexity perspectives are considered for the comparison. Moreover, to address open-set scenarios, we introduce the fusion of two OOD detectors each of which is fitted to a Gaussian mixture model by using the learned feature space (e.g. the feature embeddings and logits) from the multi-head classifica- tion network. The overall training process is achieved on two stages following a post-hoc design: the in-set classification neural network is trained first, then the learnable parameters are frozen in the evaluation mode to process the training dataset again to aggregate the embeddings and logits used to fit the GMMs for OOD detection. Moreover, a proof of concept using experimental setup is implemented to validate the operation of the introduced models and algorithms in practical scenarios. The simulation- based trained model is fine-tuned using experimentally measured radio signals. For the experimental setup, a USRP-B210 acting as a transmitter, and an edge Deepwave Air-T software-defined radio, which is equipped with the fine-tuned neural networks is used as a detector or receiver.
- [4] arXiv:2611.25353 [pdf, other]
-
Title: Distributed Fiber Monitoring on a Coherent DSP ASIC: Implementation of Longitudinal Power MonitoringTakeo Sasai, Yukinobu Nakajima, Minami Takahashi, Runa Kaneko, Masanori Nakamura, Asuka Matsushita, Fukutaro Hamaoka, Etsushi YamazakiComments: 11 pages, 19 figuresSubjects: Signal Processing (eess.SP); Optics (physics.optics)
Coherent DSP ASICs and transceivers to date report a rich set of optical link metrics, yet these remain cumulative over the link and provide no spatial localization. This paper presents the implementation design of fiber-longitudinal power monitoring (LPM), which recovers spatially resolved power profiles from received traffic, on a coherent DSP ASIC chip. We first identify the challenges specific to DSP implementation, and describe the enabling algorithm and architecture. We then present the key design considerations for retaining the estimation fidelity of the on-chip LPM during in-service operation. On this basis, we implement LPM on a coherent DSP, a 3-nm 800G low-power DSP ASIC housed in an OSFP pluggable module, which constitutes, to our knowledge, the first coherent DSP ASIC and transceiver to perform distributed fiber monitoring and data communication simultaneously. The on-chip LPM is experimentally validated over links of up to 1005 km using 400ZR+ and 800ZR+ traffic, with no measurable impact on Q factor or module power consumption. Multi-vendor interoperability is also demonstrated, where the on-chip LPM operates on signals transmitted by third-party transceivers. These results extend coherent DSPs from cumulative to spatially resolved monitoring, allowing pluggable transceivers, deployable from DCI to long-haul, to act not only as cost-efficient transmission elements but also as distributed monitoring points across diverse optical network infrastructure.
- [5] arXiv:2611.25378 [pdf, html, other]
-
Title: WiFo-D: Wireless Foundation Model Based on Discrete RepresentationsSubjects: Signal Processing (eess.SP)
Wireless foundation models have shown strong generalization capabilities across heterogeneous wireless tasks and configurations. However, most existing models rely on continuous representations, limiting ultra-low-overhead channel state information (CSI) feedback, few-shot generalization, and robustness to noise. To address these limitations, this paper proposes a wireless foundation model based on discrete representations (WiFo-D), which learns multi-granularity discrete CSI representations. It comprises a CSI encoder, multi-codebook vector quantization, and a modulation-based decoder, with the learned token maps and quantized features supporting multi-rate CSI feedback, CSI reconstruction, and diverse CSI-related downstream tasks. Extensive experiments demonstrate strong CSI feedback performance at an ultra-low latent compression ratio of 1/32 with 3 bits per token, reducing normalized mean square error (NMSE) by more than 4.7 dB compared with the strongest baseline. Furthermore, for frequency-domain prediction with limited training samples, WiFo-D improves generalization and noise robustness over continuous-representation-based methods. The results also demonstrate its superiority across downstream tasks, including beam prediction, wireless localization, and line-of-sight/non-line-of-sight identification.
- [6] arXiv:2611.25409 [pdf, html, other]
-
Title: 20 ghz longitudinal surface acoustic wave resonators in 100-nm x-cut lithium niobate on silicon carbideSubjects: Signal Processing (eess.SP)
Surface acoustic wave (SAW) resonators above 15~GHz require a fast acoustic mode that retains strong electromechanical coupling ($k^2$). This work reports longitudinal SAW (L-SAW) resonators designed for high $k^2$ in 100-nm X-cut lithium niobate (LN) on 4H silicon carbide (SiC), operating up to 21.7~GHz and reaching a $k^2$ of 8.1\%, a maximum Bode $Q$ of 70, and a figure of merit (FoM) of 5.7 at 20.24~GHz. In this thin-film waveguide, shrinking the wavelength ($\lambda$) keeps the normalized LN thickness ($h_\mathrm{LN}/\lambda$) large, sustaining both $k^2$ and acoustic confinement. Conversely, increasing $\lambda$ from 292 to 471~nm lowers $h_\mathrm{LN}/\lambda$ from 0.34 to 0.21 and raises the L-SAW phase velocity from 6.32 to 7.05~km/s, within 1.1\% of the 7126-m/s shear-wave velocity of 4H-SiC. As the velocity approaches this limit, more of the acoustic energy spreads into the non-piezoelectric SiC, and $k^2$ falls from its 8.1\% peak at $\lambda = 322$~nm to 1.7\% at $\lambda = 471$~nm. No L-SAW resonance is observed for $\lambda \geq 518$~nm ($h_\mathrm{LN}/\lambda \leq 0.19$), where the mode becomes leaky. Overall, the FoM stays within 5\% of its maximum for $h_\mathrm{LN}/\lambda$ between 0.26 to 0.31, defining a design window for L-SAW resonators in thin LN on SiC at K-band and beyond.
- [7] arXiv:2611.25556 [pdf, html, other]
-
Title: An LLM-Orchestrated Agent Framework for CKM-Assisted Network Deployment OptimizationComments: In Proc. IEEE ICCC 2026Subjects: Signal Processing (eess.SP)
Channel knowledge maps (CKMs) offer reusable, site-specific radio priors to facilitate environment-aware network planning. However, deploying heterogeneous network nodes to meet diverse performance requirements in complex environments is challenging and requires executable workflows that coordinate scene preparation, candidate-site specification, CKM construction, optimization, and validation. This paper presents a large language model (LLM)-orchestrated agent framework for CKM-assisted network deployment optimization, where an LLM translates high-level requirements into modular, typed, and reproducible planning workflows. In this framework, an LLM layer interprets intent, selects modules, and populates normalized configurations; physics-grounded modules generate scene, CKM, optimization, and validation artifacts; and a deterministic guarded controller performs validation-guided refinement. A common target-candidate-CKM-optimizer interface supports the registration and composition of compatible planning modules, with CKM backends, samplers, and solvers remaining replaceable. We instantiate the proposed framework for base station (BS) deployment, intelligent reflecting surface (IRS) deployment, and unmanned aerial vehicle (UAV) trajectory planning. An urban-block case study demonstrates automatic scene consistency checks, bounded BS/IRS deployment refinement, and collision- and radio-feasible UAV planning.
- [8] arXiv:2611.25678 [pdf, html, other]
-
Title: Lightweight Low-Frequency Non-Intrusive Load Monitoring via Karhunen-Loève Feature Distillation and Temporal Convolutional NetworksSanka Peeris, Pandula Thennakoon, Praveen Sumanasekara, Roshan Godaliyadda, Parakrama Ekanayake, Janaka EkanayakeComments: 10 pages, 6 figures. Submitted to IEEE Transactions on Smart Grid. S. Peeris, P. Thennakoon and P. Sumanasekara contributed equallySubjects: Signal Processing (eess.SP); Machine Learning (cs.LG); Systems and Control (eess.SY)
Non-Intrusive Load Monitoring (NILM) is a critical enabler for demand-side management which allows for smart grid optimization. A deployment-ready NILM solution necessitates three key features: accurate disaggregation, low-frequency sampling, and low computational cost. This paper proposes a lightweight, edge-oriented NILM framework that combines activity-boundary-aware synthetic data augmentation with Karhunen-Loève Transform (KLT) feature extraction and a Temporal Convolutional Network (TCN). The augmentation stage extracts and recombines complete or near-complete appliance activation segments so that the model can be trained without new fully labeled household aggregates. The KLT filter bank extracts decorrelated, appliance-informed temporal features which are then fed into a compact TCN along with the original aggregate for energy disaggregation. Evaluations on two standard NILM datasets, UK-DALE and REDD, at 10-s sampling and with leave-one-house-out testing, demonstrate that the proposed architecture achieves comparable or superior disaggregation accuracy relative to NILMFormer, BERT4NILM, BiLSTM, and BiGRU. The computational footprint of the framework is 31,009 parameters, the fewest of the five models compared, and on a single CPU thread it needs 2.78 ms per 1000 samples, 84-95% less than the baselines. These results demonstrate a favorable accuracy-efficiency tradeoff for low-frequency smart-meter NILM.
- [9] arXiv:2611.25726 [pdf, html, other]
-
Title: A closed-loop agentic research environment for MRI development within Siemens IDEA platformComments: 8 pages, 5 figures. Submitted to ISMRM 2027Subjects: Signal Processing (eess.SP)
Motivation: Testing new ideas in MRI can require substantial effort to learn sequence programming and reconstruction tools and to debug implementations before the scientific question can be addressed. Goals: Reduce the implementation and debugging burden of testing MRI ideas while retaining researcher control of scientific decisions. Approach: We developed an environment where LLM agents develop sequences in the native Siemens IDEA platform, evaluate them through Bloch simulation and reconstruction, and further iteratively refine them using scanner feedback. Results: As test cases, the agentic environment was used to implement and validate a GIRF-based gradient-calibration sequence and reconstruction in approximately 4 hours, and to develop and optimize a rapid 3-second multichannel calibration scan in approximately 12 hours of researcher-guided iteration.
- [10] arXiv:2611.25769 [pdf, other]
-
Title: Signed Powers-of-Two Greedy Vector Approximation of Zadoff-Chu SequencesLuiz Coelho (CEDRIC - LAETITIA, PEE/COPPE-UFRJ), Paulo Diniz (PEE/COPPE-UFRJ), Didier Le Ruyet (CEDRIC - LAETITIA)Journal-ref: IEEE Wireless Communications Letters, 2026, pp.1-1Subjects: Signal Processing (eess.SP)
This paper explores the use of sums of signed powers-of-two (SOPOT) for a representation suitable for multiplierless applications of ZC sequences, where multiplications are reduced to a prescribed number of shift-and-add operations. The SOPOT representation is achieved using a greedy algorithm that outperforms other approaches for multiplierless approximations, such as CORDIC and CSD. The proposed algorithm is based on the matching pursuits using generalized bit planes (MPGBP), and the proposed modification enables robustness to hyperparameter tuning. The resulting sequences retain, to some extent, the good autocorrelation properties of ZC, as shown in a matched-filter detection simulation. This corroborates to their use in various practical applications across mainstream cellular systems and devices requiring moderate computational cost.
- [11] arXiv:2611.25800 [pdf, html, other]
-
Title: Personalized Privacy-Aware MARL Offloading in the Edge-Cloud ContinuumComments: 6 pages, 2 figures. Accepted for presentation at IEEE CloudNet 2026Subjects: Signal Processing (eess.SP)
Task offloading in End-Edge-Cloud computing enables resource-constrained User Devices (UDs) to execute computation-intensive applications with reduced latency and energy consumption. However, existing task offloading schemes generally overlook the heterogeneous privacy requirements of users, leading to inefficient privacy-aware offloading decisions. In this paper, we propose a personalized privacy-aware task offloading framework, where the privacy requirement of each task is quantified through a User-Centric Privacy (UCP) model that considers task sensitivity, user privacy preferences, and task urgency. The task offloading problem is formulated as a Markov Decision Process (MDP) to jointly optimize the offloading ratio and the target computing server while minimizing the long-term system cost under latency, energy, and privacy constraints. To efficiently solve the resulting high-dimensional continuous optimization problem, we develop an Advantage-Weighted Actor-Critic (AWAC)-enhanced Proximal Policy Optimization (PPO)-based Multi-Agent Reinforcement Learning (MARL) algorithm under the Centralized Training and Decentralized Execution (CTDE) paradigm. Simulation results demonstrate that the proposed framework achieves lower average system cost, lower energy consumption, and higher task success rate than representative baselines under diverse privacy requirements.
- [12] arXiv:2611.25897 [pdf, html, other]
-
Title: DISC: A Deterministic Instantaneous-SINR Combiner for Uplink Massive MIMO SystemsComments: Submitted to IEEE TWC on 4th November 2026. Copyright 2026 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material, creating new works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other worksSubjects: Signal Processing (eess.SP)
Future Massive MIMO (multiple-input multiple-output) systems will fit more antennas into compact arrays, making minimum mean-square-error (MMSE) combining computationally demanding. Maximum-ratio (MR) combining serves as a cheaper alternative but performs poorly in multi-user scenarios. Existing solutions to this either mimic the MMSE combiner using second-order statistics or, as in the optimal bilinear equalizer (OBE), rely on the loose use-and-then-forget (UaTF) bound. In this paper, we propose DISC, a novel receive combiner that instead maximizes a deterministic approximation of the instantaneous signal-to-interference-plus-noise ratio (SINR), obtained by replacing the random terms in the SINR with deterministic equivalents. This approximation becomes increasingly accurate as the number of antennas grows, and it exploits useful information discarded by the UaTF bound. The computational complexity of DISC matches that of OBE, with the low-complexity variant DISC-D being even cheaper than conventional MR combining. We further prove that the SINR achieved by DISC grows unbounded with the number of antennas, even under pilot contamination, similarly to MMSE combining. Finally, we numerically demonstrate that DISC achieves a sum spectral efficiency close to that of MMSE combining under full pilot sharing, while outperforming other combiners across signal-to-noise ratio, pilot length, and spatial correlation models.
- [13] arXiv:2611.25932 [pdf, html, other]
-
Title: Coordinated Offloading: Cooperative Multi-Agent Reinforcement Learning for the Edge-Cloud ContinuumComments: 6 pages, 5 figures, 1 table. Accepted at IEEE GLOBECOM 2026Subjects: Signal Processing (eess.SP)
The increasing demand for intelligent, low-latency services in edge-cloud continuum systems poses new challenges for dynamic and efficient task offloading. We propose a Multi-Agent Reinforcement Learning (MARL) framework for distributed task offloading under partial observability, where each device offloads only a portion of its computation to a target edge server while executing the remaining portion locally (partial offloading). The problem is formulated as a Markov Decision Process (MDP), where each device independently learns an optimal policy under partially observable system states. To address the non-convexity, high-dimensional state space, and continuous action space, we develop a multi-agent Advantage-Weighted Actor-Critic (AWAC)-enhanced Proximal Policy Optimization (PPO)-based computation offloading algorithm. The proposed method combines the stable policy updates of PPO with an advantage-weighting mechanism inspired by AWAC to emphasize higher-quality actions during training. To encourage cooperation without inter-device communication, a shared global reward aligned with system cost is employed. Extensive simulations demonstrate rapid convergence and significant improvement in overall system cost compared with baselines, e.g., Deep Deterministic Policy Gradient (DDPG) and Deep Q-Network (DQN).
- [14] arXiv:2611.26038 [pdf, html, other]
-
Title: Channel-Aware Codebook Design for Task-Oriented Token Communication Over Wireless NetworksSubjects: Signal Processing (eess.SP)
This paper investigates the adaptive codebook design for task-oriented token communication over wireless networks with dynamic channels. This paper proposes the channel-aware codebook (CAC), a joint channel- and reconstruction-driven framework that reorders the labels of a frozen 14-bit VQGAN codebook. In the considered framework, a balanced geometric partition forms prefix groups, a prefix bijection places the groups on the label hypercube, and bijective suffix assignments place the codewords within each group. The problem is formulated to minimize the expected full-image perceptual distortion of the frozen decoder under a binary symmetric channel, averaged over a design set of bit error rates (BERs) and retaining simultaneous token errors. An offline algorithm alternates prefix and suffix exchanges and screens and confirms candidates on paired batches under a fixed decoder-evaluation budget. The assignment preserves the error-free reconstruction and the payload and deploys as static lookup tables without retraining. Under the binary symmetric channel at BER $0.01$, it reduces LPIPS by $12.3\%$ and improves COCO AP by $14.1\%$ ($1.72$ points) relative to the native labels, and it lowers LPIPS by a further $4.5\%$ relative to a related assignment adaptation at the same payload.
- [15] arXiv:2611.26121 [pdf, html, other]
-
Title: SEP Analysis of Fluid Antenna GSC Reception under Correlated Nakagami-m Fading and Imperfect Channel EstimationComments: 13 pages, 7 figuresSubjects: Signal Processing (eess.SP)
Channel estimation (CE) errors and correlation among the fading channels are two of the most fundamental factors degrading the performance of any multi-antenna system. This paper aims to analytically capture their joint effect on fluid antenna systems (FASs) performing generalized selection combining (GSC), i.e., combining the received signals at the best $\tilde{L}$ of the $L$ available ports using maximum ratio combining (MRC). More specifically, the FAS receiver is subject to arbitrarily correlated Nakagami-$m$ fading channels following the fully correlated Jakes' model, and employs a pilot-aided maximum-likelihood rule for CE prior to data detection. A novel tight moment-matching-based Gamma approximation is proposed for the marginal distribution of the estimated channel gains, which is then used to derive the joint statistics of the estimated correlated channels for the case of non-identical integer or half-integer fading severity parameters. Considering that the transmitter employs $M$-ary phase-shift keying or quadrature amplitude modulation, the joint statistics are further exploited to obtain exact finite-integral symbol error probability (SEP) expressions for a GSC-based FAS receiver, along with closed-form counterparts for the case of identical fading parameters. In addition, asymptotic SEP expressions at high signal-to-noise ratio are derived, analytically establishing that non-ideal CE and fading correlation degrade only the FAS's coding gain, while the diversity order is preserved and governed solely by the number of combined ports and the estimated fading severity. The presented numerical SEP investigations offer concrete design insights into the trade-offs between CE quality, fading correlation, and port selection for the GSC-based FAS receiver.
- [16] arXiv:2611.26133 [pdf, html, other]
-
Title: Absolute User Fairness for a Downlink Multiuser MIMO-NOMA SystemComments: This work has been submitted to the IEEE for possible publication. Copyright may be transferred without notice, after which this version may no longer be accessibleSubjects: Signal Processing (eess.SP)
User fairness has been widely researched in various non-orthogonal multiple access (NOMA) systems. In this paper, max-min or min-max fairness is adopted for a downlink multiuser multiple-input multiple-output NOMA system, where one multi-antenna base station (BS) transmits to multiple NOMA clusters and each cluster is composed of two single-antenna users sharing a single beam on the same time-frequency resource block. Specifically, under perfect channel state information in the transmitter (CSIT), max-min fairness is adopted to maximize the minimum achievable rate among all users. Under partial CSIT, i.e., the BS only knows the distance and channel direction information of each user, min-max fairness is adopted to minimize the maximum outage probability among all users. Under either CSIT case, to avoid solving the coupled problem, zero-forcing beamforming is designed to obtain a power control problem and then a binary search or formula based algorithm is proposed. Theoretical analyses reveal that under either CSIT case, the problem can be guaranteed to have a unique solution and the proposed algorithm is computationally efficient due to only simple closed-form expressions involved.
- [17] arXiv:2611.26137 [pdf, other]
-
Title: Reading Across Space, Time and Frequencies: Tensor Decomposition of SEEG Signals to Support Epileptogenic Zone DefinitionFederico Mason, Matteo Neri, Maria Fratello, Julia Makhalova, Fabrice Bartolomei, Giovanni Petri, Christian G. BenarSubjects: Signal Processing (eess.SP)
Stereo-electroencephalography (SEEG) is an invasive technique used in the presurgical evaluation of drug-resistant epilepsy to support the localization of the Epileptogenic Zone (EZ). SEEG data are particularly complex to analyze, as they include hundreds of components, named channels, each associated with a different brain site, with interactions varying in both time and frequency. In this study, we propose a new framework for SEEG analysis, based on the Canonical Polyadic (CP) decomposition, to extract the latent modes underlying seizure generation. We model SEEG signals as a third-order tensor (channel x time x frequency), and factorize the tensor into rank-1 factors. We then exploit the channel-specific factor as an unsupervised feature set for identifying the EZ. The framework is validated on SEEG-recorded seizures from 35 patients with drug-resistant epilepsy. The results show that the joint combination of multiple frequency bands significantly enhances classification performance.
- [18] arXiv:2611.26142 [pdf, html, other]
-
Title: Transition Dynamics Matter: Hidden Markov Scoring of Epileptogenic Channels in SEEGGiulia Tacconi, Federico Mason, Lorenzo Ferri, Elena Pasini, Lidia Di Vito, Matteo Martinoni, Lara Alvisi, Anna Scarabello, Luca Zanuttini, Francesca Bisulli, Andrea ZanellaSubjects: Signal Processing (eess.SP); Medical Physics (physics.med-ph)
Stereo-electroencephalography (SEEG) represents the most effective technique for the detection of the Epileptogenic Zone (EZ) in drug-resistant epilepsy. This procedure produces a multi-dimensional signal with hundreds of components, named channels, each associated with a different brain location. In this work, we propose a joint Hidden Markov Model (HMM), in which each SEEG channel is described by a latent Markov chain, with states shared across the network and channel-specific transition probabilities. Accordingly, we develop a detection algorithm that identifies the channels belonging to the EZ by quantifying, through the Kullback-Leibler divergence, the difference between their transition probabilities in the inter-ictal period and at seizure onset. We evaluate our framework in SEEG recordings from 30 patients with drug-resistant epilepsy and compare it with a benchmark that operates in the same observation space but does not consider a latent-state structure. Compared against medically validated EZ definitions, our study shows that modeling channel-specific temporal transitions can improve EZ localization and offer a data-driven framework to investigate the latent dynamics underlying seizures.
- [19] arXiv:2611.26258 [pdf, html, other]
-
Title: Temporal Contrastive Alignment of EEG and Naturalistic Video Only Beats Linear Decoding On In-Sample StimuliSubjects: Signal Processing (eess.SP); Neurons and Cognition (q-bio.NC)
Neural tracking, the temporal alignment between a continuous stimulus and the elicited brain response, enables decoding of attention from electroencephalography (EEG), yet its application to naturalistic video remains largely unexplored and mainly relies on linear canonical correlation analysis (CCA). We investigate whether a temporal contrastive model, pairing pretrained VideoMAE features with a lightweight EEG encoder trained on a correlation-based InfoNCE objective, can strengthen EEG-video alignment. Benchmarked against CCA under three generalization regimes, the model outperforms CCA across subjects (leave-one-subject-out), matches it within the stimulus manifold (leave-video-segment-out), and falls behind CCA when extrapolating to entirely novel videos (leave-one-video-out). Because this advantage appears only for seen stimuli, we probe its basis: a sham control shows that a high accuracy does not certify genuine alignment, as a synthesized correspondence fit to the seen stimuli can reproduce it without genuine temporal alignment in the training pairs. This does not preclude practical use: on a downstream attention-decoding task with seen stimuli, the model remains better than CCA since the decoding only requires a strong statistical coupling, regardless of whether that coupling is physiologically meaningful. Using a group-averaged EEG reference, we further show that genuine, generalizable coupling is nonetheless learned, only buried in the noise of single-subject EEG. These results indicate that progress hinges on data scale, diversity, and single-subject denoising.
- [20] arXiv:2611.26270 [pdf, html, other]
-
Title: Chance-Constrained Joint Beam Hopping Scheduling and Bandwidth Allocation for LEO SatellitesOlfa Ben Yahia, Jean-Francois Frigon, Brunilde Sanso, Stephane Martel, Guillaume Mantelet, Antoine Lesage-Landry, Gunes Karabulut KurtSubjects: Signal Processing (eess.SP); Systems and Control (eess.SY)
Service level agreement (SLA) reliability is a key requirement in multibeam satellite systems, where operators must satisfy heterogeneous and time-varying traffic demands using limited resources. Beam hopping (BH) enables flexible resource allocation by activating only a subset of beams at each time slot. However, most existing BH optimization methods rely on predicted channel conditions and do not explicitly control the probability of SLA violation under spectral efficiency uncertainty. This work introduces an SLA-aware chance-constrained BH optimization model for joint beam scheduling and bandwidth allocation. The proposed formulation maximizes the expected delivered traffic while ensuring that each cell demand is satisfied with a target reliability level. By modelling the spectral efficiency prediction error as Gaussian, the probabilistic SLA constraints are reformulated as deterministic second-order cone constraints. Evaluation results demonstrate that the proposed method maintains the target SLA reliability under increasing traffic load and spectral efficiency uncertainty, while maintaining fairness and efficient resource allocation.
- [21] arXiv:2611.26393 [pdf, html, other]
-
Title: Sensor-Free Shockable Rhythm Classification During CPR Using the Existing ECG Channel of Public-Access Automated External DefibrillatorsSubjects: Signal Processing (eess.SP)
Automated external defibrillators (AEDs) are public-access devices intended for use by lay rescuers, yet they must pause chest compressions to analyze the rhythm because compressions corrupt the electrocardiogram (ECG). Removing this pause requires a method that uses the hardware already present in the device, fits within its processor budget, and meets the rhythm-recognition performance required for defibrillator approval. Existing approaches rely either on an auxiliary sensor, which entails a hardware redesign, or on an additional network that reconstructs the ECG before classification. This work classifies the compression-contaminated ECG directly, using only the existing ECG channel of the AED. A cascade of three compact convolutional networks trained with different class ratios verifies each decision with a second network biased toward the proposed outcome; disagreement yields an indeterminate outcome. Evaluated on an independent manufacturer-provided cohort of 464 subjects over 15 train-validation partitions, the cascade satisfies both adjudicable American Heart Association criteria under all six noise conditions at an eight-second window and a signal-to-artifact ratio of 0 dB, with sinus-rhythm specificity of 99.20-99.43% and other non-shockable specificity of 95.48-96.20%, leaving only 1.07-1.48% of segments indeterminate. The criteria are not met at -3 dB. Single networks matched in parameter count and inference cost do not reproduce this performance. Each decision evaluates two of the three networks and requires 6.68 ms on a single thread of a general-purpose processor, and eight-bit quantization reduces the model to 1.45 MiB without changing any compliance verdict. Thus, rhythm analysis during ongoing CPR can be implemented with the ECG hardware already present in public-access AEDs, without auxiliary sensing hardware, subject to validation on the target device.
- [22] arXiv:2611.26442 [pdf, html, other]
-
Title: Decision Fusion in Wireless Networks: A Tutorial and Survey of Classical and Learning-Based MethodsComments: 36 pages, 6 figures, journalSubjects: Signal Processing (eess.SP)
Decision fusion combines local decisions from distributed sensors to support reliable detection of phenomena in wireless networks. The effectiveness of decision fusion depends on sensing accuracy, reporting-channel reliability, and available communications and computation resources. This article surveys decision-fusion techniques for wireless sensor networks and emerging communication systems, bringing together their theoretical foundations and practical implementation requirements. We review network architectures and hard, soft, and quantized reporting strategies, and examine statistical, learning-based, game-theoretic, and heuristic fusion approaches. Advanced fusion techniques, including relay-aware combining, link-adaptive regeneration, branch selection, and likelihood approximations, are discussed with emphasis on their assumptions, information requirements, and limitations. The survey examines tradeoffs among detection performance, computational complexity, communications overhead, energy consumption, and latency. The tradeoffs are considered across representative sensing applications and emerging technologies, including massive machine-type communication, ultra-reliable low-latency communication, uncrewed aerial vehicle-assisted networks, and the massive Internet of Things. The reviewed literature highlights the importance of jointly considering sensing quality, reporting reliability, and fusion processing when selecting a suitable strategy. Finally, we identify research directions in resource-aware optimization, uncertainty calibration, and robust learning-based fusion under heterogeneous data, missing reports, and changing wireless conditions. Particular attention is given to convergence, generalization, and the costs of training and inference.
- [23] arXiv:2611.26512 [pdf, html, other]
-
Title: Decision Age: Information Freshness under Fusion Rules in Multi-Sensor Command and ControlSubjects: Signal Processing (eess.SP)
Command and control systems decide automatically on evidence that a fusion rule requires from several sensors. The Age of Information measures how current a datum is when it reaches the fusion node, but existing multi-source results treat additional sources as redundancy: the age at the monitor falls as sources are added. This paper addresses the opposite regime. When a rule demands corroboration the decision waits for the slowest required source, so the binding quantity is a maximum, not a minimum, of source ages. We define the decision age of an arbitrary monotone fusion rule, obtain its distribution in closed form for periodic and event-driven sources, and show that the mean decision age grows linearly with the corroboration level, from the floor of the age sawtooth to its peak. We add a dominance criterion that identifies which sensor governs decision freshness, and a measurement methodology for fielded systems. Freshness becomes a verifiable requirement, and adding sensors is shown to make automatic decisions older rather than fresher.
New submissions (showing 23 of 23 entries)
- [24] arXiv:2611.25625 (cross-list from cs.LG) [pdf, html, other]
-
Title: QUBO-based feature selection for MRI undersampling pattern designComments: 20 pages, 3 figuresSubjects: Machine Learning (cs.LG); Signal Processing (eess.SP)
We formulate undersampling mask design for Cartesian magnetic resonance imaging (MRI) as a feature subset selection problem. Each sampling line is represented by a binary feature indicating whether it is included in each candidate mask. Using candidate masks and their training-set evaluations, we estimate mutual information (MI) and conditional mutual information (CMI). These estimates and a sampling-count penalty define a quadratic unconstrained binary optimization (QUBO) model. The model is solved using simulated annealing and quantum annealing to determine a feature subset and its corresponding sampling mask. We assess the selected subsets on the test set through zero-filled reconstruction in Fourier undersampling simulations using fastMRI knee magnitude reference images, with patient-disjoint training and test sets. At all six tested target sampling rates from 15% to 40%, the resulting reconstructions have lower mean root mean square error than those obtained with four conventional sampling methods. They also have higher mean structural similarity than those obtained using univariate mutual-information ranking at all tested rates. These results support information-theoretic feature selection as an approach to MRI undersampling mask design under the tested zero-filled simulation settings.
- [25] arXiv:2611.25756 (cross-list from cs.IT) [pdf, html, other]
-
Title: Self-Supervised Operator Learning for Channel Estimation in Fluid Antenna SystemsComments: 6 pages, 4 figures. Submitted to an IEEE conferenceSubjects: Information Theory (cs.IT); Signal Processing (eess.SP)
Fluid antenna systems (FAS) have emerged as a promising technology for enhancing wireless communications through adaptive antenna repositioning. Recently, Fourier neural operators (FNOs) have shown strong potential for FAS channel estimation owing to their ability to learn resolution-invariant mappings and generalize across different port densities. However, existing operator learning approaches require densely sampled channel realizations as training labels, whereas a practical receiver only collects noisy pilot observations from a small subset of ports. This mismatch between training requirements and practical channel acquisition poses a major obstacle to deployment. To address this challenge, we develop a self-supervised framework that trains the FNO-based estimator directly from noisy sub-sampled pilots without any labeled channel fields. The proposed objective combines a pilot splitting strategy, which creates supervision at ports hidden from the model input, with a correction based on Stein's unbiased risk estimate (SURE) that eliminates the measurement noise bias. We prove that this objective admits the same minimum mean square error (MMSE) solution as the ideal supervised criterion. Since the loss relies solely on received pilots, the estimator can be continually adapted online on the live pilot stream. Simulation results show that the self-supervised FNO achieves accuracy comparable to supervised training, generalizes zero-shot to unseen port resolutions, and rapidly recovers from distribution shift via online adaptation.
- [26] arXiv:2611.25831 (cross-list from cs.LG) [pdf, html, other]
-
Title: Beyond Alignment: Whitening Geometry, Profile Synthesis, and Heterogeneous Ensembles for Cross-Environment WiFi Action RecognitionComments: 15 pages, 3 figures, 4 tables. Preprint of a manuscript submitted to IEEE Transactions on Mobile ComputingSubjects: Machine Learning (cs.LG); Signal Processing (eess.SP)
Cross-environment deployment is the central obstacle to practical WiFi sensing: a pipeline reaching 0.89+ top-1 in its training environments collapses to 0.408 under Leave-One-Domain-Out (LODO) evaluation with only K-shot labels. A ceiling analysis shows the collapse is not an alignment problem: the action signal is strong across seen environments (about 0.89 K10), environment identity is decodable from the encoder (0.965), and the failure concentrates in the encoder's extrapolation to the novel environment. A pre-registered falsification chain confirms that "align the domains" methods--training-time alignment, readout-level prototype alignment, adversarial confusion, and covariance/subspace alignment (CORAL -3.8pp, subspace -8.3pp)--degrade performance, and that source-domain statistics anti-correlate with cross-environment contribution. The effective levers lie elsewhere; we combine them into Dec-H: (i) closed-form source within-class whitening, provably re-geometrizing the readout's L2 regularization (+8.6pp, zero cost); (ii) profile-mixing augmentation (+1.0pp); (iii) few-shot adaptation with alternating whitening refit (+1.5pp; fixed preconditioners are 20/20 negative); (iv) heterogeneous decision ensembles (+1.7pp); homogeneous scaling and source-weighted stacking are saturated or harmful. Dec-H reaches 0.560 K10 (0.812 K40), +15.2pp over the baseline, consistent across all four environments and seeds; all thirteen configurations were pre-registered with frozen criteria.
- [27] arXiv:2611.25968 (cross-list from cs.IT) [pdf, html, other]
-
Title: Feedback-Free FDD Cell-Free Massive MIMO With Sparse AP Activation and Dual-Mode PrecodingComments: 5 pages, 3 figures. Submitted to IEEE Transactions on Vehicular TechnologySubjects: Information Theory (cs.IT); Signal Processing (eess.SP)
This correspondence enables practical frequency-division duplex (FDD) operation in cell-free massive MIMO via two mechanisms. Sparse access-point (AP) activation reduces downlink pilot overhead to enable coherent detection, while angle-reciprocity precoding steers beams using only uplink-estimated angles, eliminating channel feedback entirely. A load-adaptive dual-mode optimizer further enhances performance: closed-form signal-to-leakage-and-noise-ratio precoding for high-load rate boosting, or semidefinite relaxation for low-load power minimization. Simulations verify its superiority over existing feedback-free FDD solutions in spectral and energy efficiency, and its robustness to angle errors, reciprocity mismatch, and multipath.
- [28] arXiv:2611.25977 (cross-list from eess.AS) [pdf, html, other]
-
Title: Blind tests of a simple-body model of submarine target strength on open benchmarksComments: 17 pages, 6 figures, 9 tables, Data: this https URLSubjects: Audio and Speech Processing (eess.AS); Sound (cs.SD); Signal Processing (eess.SP)
Finite-element and boundary-element models of a submarine's sonar echo need a mesh and a solver, and their cost grows quickly with frequency. SSR-Sim takes an analytical route. The hull is cut into a few simple bodies, the echo of each is a closed-form result of geometrical acoustics or of the Kirchhoff approximation, the surface enters through a plane-wave reflection coefficient, and the echoes add in energy. This paper describes the method, including the reduction of a hull and its sail to simple bodies, and tests it on three open cases. (1) The reflection coefficient of a 50 mm rubber tile on steel plating, computed with finite elements by Yang et al., is reproduced to 0.02 rms or better at incidence angles from 0° to 75° once the layer is allowed to refract. (2) For the rigid full-scale BeTSSi submarine at 1 kHz, a prediction fixed before the reference was opened agrees with the boundary-element results of Venås and Kvamsdal to within 0.7 dB in every aspect sector from the bow to 150°; in the stern sector it is 5 dB high. (3) For a 1:15 BeTSSi model with 10 mm tiles measured in a lake, a second blind prediction is within -3.1, +1.4 and -0.5 dB of the measured level over aspects 0-120° at 5, 15 and 20 kHz, against -3.1, +3.1 and +3.5 dB for the Kirchhoff-finite-element computation published with the measurement. At 10 kHz, on a resonance of the tile, the prediction is 6.9 dB low, and in the bow sector it is 7 dB or more low at every frequency. The tests cover the specular component and the coating model only. The failures show what such a model lacks: edge diffraction and the returns that set the off-beam floor of a real model.
- [29] arXiv:2611.26023 (cross-list from cs.CR) [pdf, html, other]
-
Title: Detecting Formation-Preserving GNSS Spoofing of UAV Swarms through Transmitter Parallax in Pseudorange and DopplerComments: 16 pages, 7 main figures, 7 main tables. Includes supplementary appendicesSubjects: Cryptography and Security (cs.CR); Signal Processing (eess.SP)
Swarms of unmanned aerial vehicles (UAVs) detect Global Navigation Satellite System (GNSS) spoofing by checking the drones' reported positions against their known formation. Formation-preserving spoofing evades this check: in GNSS-WASP, one transmitter per satellite, a few kilometres away, shifts the whole swarm while keeping its formation. We show that such an attack still leaves a trace below the position level. Because the transmitters are close, each drone sees them under a slightly different angle, and this parallax bends the signal arrival times across a 90 m swarm by 20-40 cm for a transmitter at 5 km, whatever delays the attacker applies. We propose a likelihood ratio test that reads this trace from the pseudoranges and Doppler every drone already measures, with all clocks and attacker delays unknown and no extra antennas. In simulations with real GPS and Galileo orbits and white receiver noise, 16 drones detect GNSS-WASP at 5 km in 13-60% of 10 s windows in straight flight and in every window during a formation turn with a precise Doppler receiver, where the countermeasure proposed with GNSS-WASP detects at most 5%. The same test also catches single-transmitter and ground-array spoofing.
- [30] arXiv:2611.26532 (cross-list from physics.optics) [pdf, other]
-
Title: From power to spectrum: an off-the-shelf photodiode turned into a spectrometer by its bias aloneComments: 15 pages, 9 figures, 5 tablesSubjects: Optics (physics.optics); Image and Video Processing (eess.IV); Signal Processing (eess.SP); Systems and Control (eess.SY)
Computational spectrometers replace dispersive optics with detectors whose spectral response can be varied, but every reported single-detector device obtains this response through dedicated fabrication: two-dimensional heterostructures, graded epitaxial junctions, organic stacks, filters or photonic circuits. We demonstrate a generic, fabrication-free alternative in which a standard commercial photodiode becomes a spectrometer through its electrical drive alone, requiring no modification of the device, no optical filter and no modulation of the light. The mechanism is bias-modulated optoelectronic chromatic dispersion (OED): light of different wavelengths is absorbed at different depths, and a sinusoidal bias changes how carriers generated at each depth are collected, so that every spectrum is encoded in the shape of the photocurrent waveform over one bias period. Applied to an unmodified InGaAs PIN photodiode at room temperature, a 10 us waveform decoded by a neural network reconstructs spectra of one to three laser lines across the 80 nm of the telecom C+L band, locating the lines to 0.3-0.4 nm, counting them in about 99% of measurements and identifying their power levels, with line shifts tracked to tens of picometres. Because OED relies on universal semiconductor transport physics, the approach is material-agnostic and applies in principle to any off-the-shelf photodiode, turning detectors already deployed as power monitors into low-cost spectral sensors.
Cross submissions (showing 7 of 7 entries)
- [31] arXiv:2602.14191 (replaced) [pdf, html, other]
-
Title: Robust SAC-Enabled UAV-RIS Assisted Secure MISO Systems With Untrusted EH ReceiversSubjects: Signal Processing (eess.SP)
Secure downlink transmission in UAV-mounted reconfigurable intelligent surface (RIS)-assisted multi-user multiple-input single-output (MISO) networks is challenged by imperfect channel state information (CSI), internal eavesdropping by untrusted energy-harvesting receivers (UEHRs), discrete RIS phase shifts, and nonlinear energy harvesting (EH). This paper jointly optimizes the UAV hovering position, base station (BS) transmit power allocation, and discrete RIS phase shifts to maximize the worst-case secrecy energy efficiency (WCSEE) under bounded CSI uncertainty and nonlinear EH constraints. The resulting problem is a highly non-convex mixed-integer fractional optimization problem with tightly coupled spatial, power, and RIS variables. To address these challenges, we develop two complementary solution frameworks. First, for perfect CSI and continuous RIS phase shifts, we derive an analytical benchmark based on block coordinate descent (BCD) and successive convex approximation (SCA). Second, for the general robust setting with discrete phase quantization and CSI uncertainty, we propose a customized soft actor-critic (SAC) framework that incorporates continuous-to-discrete action projection, twin critics to mitigate value overestimation, and a feasibility-aware reward mechanism. Numerical results demonstrate that the proposed SAC policy achieves up to $28\%$, $16\%$, and $6\%$ higher WCSEE than the SCA benchmark, DDPG, and TD3, respectively, while exhibiting faster convergence and maintaining stable performance across varying user densities, UEHR counts, and CSI uncertainty levels.
- [32] arXiv:2602.23924 (replaced) [pdf, html, other]
-
Title: Modeling and Link Budget Feasibility Analysis of Secure LoRa-Based Peer-to-Peer Communication for Short-Range Tactical NetworksSubjects: Signal Processing (eess.SP); Audio and Speech Processing (eess.AS)
Short-range reliable and secure communication is a major priority in the tactical, military and disaster response settings where the traditional communication infrastructure is either off-line or prone to interception. Current VHF/UHF radios and software-defined radios are popular but large-sized devices and require lots of power, making them not suitable to be used as lightweight wearable devices with seamless hand-free use. In this paper, the design and theoretical framework of a miniature, LoRa based encrypted intercommunication device that can be used in secure field communication over a range of 1-1.5km and under line-of-sight conditions is provided. The suggested system consists of a voice-activated acquisition block, digital audio compression, an embedded microcontroller processor, and AES-128 encryption followed by a low-power transmission via the LoRa protocol. Through the ability of chirp spread spectrum modulation to utilize the long-range and low-energy properties, the system is guaranteed reliable communications coupled with low power consumption and low electromagnetic footprint. The theoretical analysis of the proposed communication range is justified using a link-budget that justifies the practicability of the communication range in the real propagation conditions. This architecture focuses on infrastructural agnosticism, peer-to-peer security as well as wearable ergonomics. The given scheme shows the possibilities of LoRa technology in the scope of other traditional IoT telemetry, and it can be further extended to include secure tactical voice communication platforms.
- [33] arXiv:2605.10145 (replaced) [pdf, html, other]
-
Title: A Generative AI-Enhanced Digital Twin Framework for Proactive Interference Management in Hybrid Near/Far-Field Wireless SystemsComments: 16 pages, 7 figures, 6 TablesJournal-ref: IEEE Transactions on Wireless Communications 2026Subjects: Signal Processing (eess.SP)
The applications of Digital Twins (DTs) and Generative AI (GenAI) have demonstrated significant potential for modeling and learning-based wireless communications. However, their joint potential for proactive wireless system design remains largely underexplored, particularly in extremely large-scale multiple-input multiple-output (XL-MIMO) networks characterized by hybrid near-field (NF) and far-field (FF) propagation. In this work, we propose an integrated GenAI-enhanced DT framework for proactive interference management in dynamic indoor environments. The DT constructs a site-specific virtual replica of the deployment, enabling geometry-aware modeling of blockage and hybrid NF/FF propagation. Conditioned on the DT environment and historical network states, GenAI generates multiple plausible future channel and blockage realizations, which are incorporated into regime-aware proactive beamforming. Extensive Sionna ray-tracing simulations across open and cluttered indoor environments, including different user equipment (UE) speeds, demonstrate improved interference suppression, signal-to-interference-plus-noise ratio (SINR), and prediction accuracy compared with reactive and deterministic DT-based benchmarks.
- [34] arXiv:2607.17591 (replaced) [pdf, html, other]
-
Title: Tensor-Based Dynamic Channel Estimation for mmWave Movable Antenna MIMO SystemsSubjects: Signal Processing (eess.SP)
This paper investigates the dynamic channel estimation algorithm in mmWave movable antenna (MA) multiple-input multiple-output (MIMO) systems. To achieve highly accurate channel estimation, we propose a tensor decomposition-based channel estimation algorithm. First, by leveraging the path response model and utilizing the intrinsic sparsity of mmWave channels, the channel corresponding to MA pairs at the base station and mobile station is transformed into a superposition of channels from sparse paths. Next, the received signal is constructed as a fourth-order tensor to fully capture the high-dimensional structural information of the MA MIMO channel. Then, two tensor decomposition schemes are adopted to extract the factor matrices, and our analysis reveals that the uniqueness of the decomposition can be guaranteed in our model. Subsequently, the propagation loss, frequency offset, angle of arrival/departure, and time delay are obtained based on these factor matrices and the channel matrix can be rebuilt. Additionally, Cramér-Rao bound (CRB) is also derived as a performance evaluation standard, proving that the proposed algorithm achieves a higher estimation accuracy and nearly approaches this minimum bound. Moreover, normalized mean square error (NMSE) is selected as the evaluation metric for estimation accuracy. Finally, simulation results reveal a notable reduction in the estimation error of the proposed algorithm when compared to the baseline algorithms, confirming its estimation advantage.
- [35] arXiv:2607.17862 (replaced) [pdf, html, other]
-
Title: Massive MIMO-OFDM ISAC for Sparse ISAR Imaging: Joint Power and Subcarrier AllocationSubjects: Signal Processing (eess.SP)
This paper investigates a massive multiple-input multiple-output (mMIMO) orthogonal frequency-division multiplexing (OFDM) framework for integrated sensing and communication (ISAC) with inverse synthetic aperture radar (ISAR) imaging. A dual-function architecture combines communication precoding and dedicated sensing beamforming to support simultaneous downlink communication and ISAR imaging. To address incomplete measurements arising from intermittent pilot transmission and sparse sensing-subcarrier activation, an adaptive reweighted two-dimensional alternating direction method of multipliers (ADMM) algorithm is developed for sparse ISAR reconstruction. A joint resource-allocation framework is formulated to optimize communication-subcarrier assignment, sensing-subcarrier selection, and transmit power allocation under communication quality-of-service (QoS) and sensing constraints. Exploiting channel hardening, analytical full-band benchmarks based on statistical channel state information (CSI) are derived for maximum-ratio (MR) and zero-forcing (ZF) precoding, while a soft actor-critic (SAC) algorithm is developed for sparse-sensing resource allocation. Numerical results demonstrate that the adaptive ADMM algorithm improves ISAR reconstruction accuracy over conventional methods, while the SAC-based design achieves higher sum spectral efficiency than both the analytical full-band ISAC benchmarks and the deep deterministic policy gradient (DDPG) baseline under both MR and ZF precoding, while satisfying the communication QoS and sensing constraints.
- [36] arXiv:2608.29020 (replaced) [pdf, html, other]
-
Title: A Continuous Payload-Bearing Discrete Multitone Modulation Framework for Fiber-Optic Integrated Sensing and CommunicationComments: This manuscript has been submitted for consideration for publicationSubjects: Signal Processing (eess.SP)
A key challenge in fiber-optic integrated sensing and communication (ISAC) is to make the payload-bearing waveform itself serve both functions without a separate sensing waveform or sensing-only silent interval. We propose a continuous discrete multitone (DMT) framework with two waveform modes, in which the same payload-bearing waveform supports forward intensity-modulation/direct-detection (IM/DD) communication and backward distributed acoustic sensing (DAS). A unified phase-sensitive optical time-domain reflectometry ({\phi}-OTDR) model represents distributed Rayleigh backscattering as a finite-memory sensing multipath channel. It shows that conventional pulse-and-wait {\phi}-OTDR requires a round-trip-time-scale silent interval to isolate successive returns, while nonzero off-peak samples in practical matched-filter (MF) pulse compression cause spatial intersymbol interference (ISI). Continuous DMT instead retains superposed returns and separates range-cell contributions through known-waveform channel reconstruction. Cyclic-prefix DMT (CP-DMT) uses a full-memory CP and one-tap frequency-domain equalization (FDE); under sufficient-CP and ideal full-bin inversion conditions, it eliminates MF-correlation-induced spatial ISI. Cyclic-prefix-free DMT (NoCP-DMT) applies regularized least squares (LS) to the long-memory linear convolution, avoiding the CP at higher receiver complexity. Experiments over a 10-km fiber link localize a 600-Hz disturbance applied by a piezoelectric transducer (PZT) near 5.071 km and recover gauge-differential phase with correlations of 0.989 and 0.987 for CP-DMT and NoCP-DMT, respectively. At 2- and 1-V PZT drive levels, the ten-record localization standard deviations are 0.16/0.11 m and 0.18/0.38 m, respectively. The corresponding IM/DD error vector magnitude values range from 4.80% to 5.00%, with no bit errors observed.
- [37] arXiv:2605.02989 (replaced) [pdf, other]
-
Title: Information Theory and Statistical LearningSubjects: Information Theory (cs.IT); Signal Processing (eess.SP); Machine Learning (stat.ML)
This manuscript contains preprint of a chapter in the forthcoming third edition of Cover and Thomas's Elements of Information Theory, posted with permission from Wiley. For feedback, please contact abbas@ee.this http URL.
Learning and information theory intersect in both model training and the characterization of fundamental performance limits. This manuscript provides a concise and accessible treatment of the first intersection, requiring only basic background in information theory and statistics at the senior undergraduate / first-year graduate level. End-of-chapter exercises make the material well suited for classroom use as well as self-study.
The chapter focuses on the role of divergence measures in model training, with examples ranging from linear and logistic regression to autoregressive models, variational autoencoders, diffusion models, and score-based models. In particular, the treatment of the generative diffusion model provides a more systematic and explicit derivation than is typical in the literature. - [38] arXiv:2605.12026 (replaced) [pdf, html, other]
-
Title: Spectral Vision Transformer for Efficient Tokenization with Limited DataAlexandra G. Roberts, Maneesh John, Jinwei Zhang, Dominick Romano, Mert Sisman, Ki Sueng Choi, Heejong Kim, Mert R. Sabuncu, Thanh D. Nguyen, Alexey V. Dimov, Pascal Spincemaille, Brian H. Kopell, Yi WangComments: 30 pages, 9 figures, 14 tables. v2: expanded experiments (MedMNIST3D, natural images, object shapes) and block spectral extensionSubjects: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Signal Processing (eess.SP)
Vision transformers tokenize an image into patches, which couples the token count to resolution and requires the embedding to be learned from data. We propose a spectral vision transformer architecture for efficient tokenization in the limited-data regime, with an emphasis on medical imaging. Rather than using spatial patches, we tokenize into coefficients of a fixed basis. This approach defines each token as one spectral mode, and the sequence length is chosen rather than implied by resolution, reducing the cost of attention. Our proposed spectral ViT shows superior or comparable performance where data is scarce in public medical imaging benchmarks, natural images, object shapes, simulations, and protected clinical data. We release our code at: this https URL.
- [39] arXiv:2607.22234 (replaced) [pdf, html, other]
-
Title: Finite-Support Structure of Optimal i.i.d. Inputs for Finite-Memory Poisson ChannelsComments: 34 pages, 7 figuresSubjects: Information Theory (cs.IT); Signal Processing (eess.SP)
Discrete-time Poisson channels with finite intersymbol interference provide a natural model for direct-detection optical links in which multipath memory and signal-dependent shot noise appear simultaneously. Under peak and average optical-intensity constraints, we study the stationary mutual-information-rate optimization under an independent and identically distributed (i.i.d.) input restriction. We prove that every input distribution maximizing the stationary mutual information rate within the i.i.d. input class has finite support. The proof is carried out directly on the entropy rate of the continuous-state hidden Markov output process induced by the finite-memory channel. We first establish a filtering-forgetting estimate whose constants are uniform over all admissible i.i.d. input laws. We then derive the entropy-rate first variation, construct a holomorphic extension of the corresponding influence function, and combine the Karush-Kuhn-Tucker condition with a supralinear growth argument.
- [40] arXiv:2608.07385 (replaced) [pdf, html, other]
-
Title: Omni-modal decomposition autoencoders learn full-stack wearable disentangled representationsIoannis Ziogas, Ensieh Khazaei, Bilal Taha, Aamna Al Shehhi, Ahsan H. Khandoker, Leontios J. Hadjileontiadis, Dimitrios HatzinakosComments: 14 pages, 7 figures, 7 tablesSubjects: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Audio and Speech Processing (eess.AS); Signal Processing (eess.SP); Machine Learning (stat.ML)
Learning disentangled representations is key for general-purpose multi-modal wearable models, yet existing approaches rarely combine classification, disentanglement, fusion, and generation of heterogeneous multi-modal time series in a single model. We introduce Omni-modal Variational Decomposition Autoencoders (OmniDecVAEs), a framework that learns multi-purpose representations from arbitrarily many modalities with a single shared encoder-decoder. OmniDecVAEs extend DecVAEs by learning time-frequency latent subspaces per modality through a multi-view self-supervised decomposition loss and a component-wise conditional decoder. On the omni-modal human activity recognition (HAR) HARWE dataset (up to thirty channels) and the public UCI HAR and WESAD datasets, OmniDecVAEs learn the most disentangled representations for unseen subjects (DCI disentanglement up to 3.0 percentage points (pp) higher). On HARWE, they improve subject-dependent HAR and identity recognition by 1.1 pp and 5.8 pp, and outperform supervised end-to-end models by +31.6 pp and +22.3 pp on unseen subjects of multi-device datasets. Moreover, their generative capacity improves reconstruction mean squared error over multi-modal VAEs (34-74% improvement) and their synthetic data improve affect recognition on unseen subjects (2 pp). With 4.1M parameters, invariant to the number of modalities (21xfewer than a multi-branch VAE), OmniDecVAEs are scalable, single-model full-stack learners for wearable representations.