IEEE Access最新文献

{"title":"Clean-Anchor Alignment: Toward Robust and Stable Graph Neural Networks","authors":"Joseph Cho;Chaoning Zhang;Sung-Ho Bae;Sungyoung Lee","doi":"10.1109/ACCESS.2026.3658657","DOIUrl":"https://doi.org/10.1109/ACCESS.2026.3658657","url":null,"abstract":"Graph neural networks (GNNs) are sensitive to small, plausible changes in structure and node features that frequently occur in real data. We formalize robustness as neighborhood stability: the model should yield the same output in a reasonable disturbance range around each clean graph and propose a clean anchor alignment (CAA) framework. A teacher provides a distribution on the clean input, while a student learns on perturbed views and is softly aligned to that clean anchor through a confidence gate and a short schedule that increases alignment after the task head has settled. CAA is architecture agnostic, requires only one additional clean forward per batch, and uses an exponential-moving-average (EMA) teacher without gradient updates, which keeps optimization stable and the cost low. In our implementation, this extra teacher pass keeps total training time in the range of seconds to at most a few tens of seconds on a single GPU. On five standard benchmarks (Cora, Citeseer, PubMed, MUTAG, PROTEINS) and multiple non-adversarial perturbation families, CAA improves Robust@<inline-formula> <tex-math>$rho {=}0.10$ </tex-math></inline-formula> by about+2.7–3.9 percentage points over a strong supervised baseline and reduces a prediction-dispersion metric by approximately 29.3%. Under feature-space FGSM attacks, CAA further mitigates the accuracy drop, with relative reductions between 2.5% and 13.1% as the perturbation strength increases. Clean accuracy remains competitive, indicating that the gains in robustness do not come at the expense of performance on unperturbed data.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"14 ","pages":"22274-22288"},"PeriodicalIF":3.6,"publicationDate":"2026-01-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11366644","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146175585","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Spatial Temporal Eye-Tracking Indicators of Novice Programmers’ Skill","authors":"Zubair Ahsan;Unaizah Obaidellah","doi":"10.1109/ACCESS.2026.3659010","DOIUrl":"https://doi.org/10.1109/ACCESS.2026.3659010","url":null,"abstract":"As Large Language Models (LLMs) become more common in education, there is growing concern about how to accurately assess learners’ skills, particularly in programming, where understanding cognitive strategies is essential. Traditional measures often overlook the sequential and temporal nature of how students read and comprehend code. This study presents a deep learning-based framework that models spatial-temporal eye-tracking data using long short-term memory (LSTM) and Transformer architectures to classify programming skill levels. We examine both binary (high vs. low) and ternary (high, average, low) labelling strategies to assess their impact on classification performance and instructional relevance. Findings from a stimulus-wise evaluation show that binary classification leads to better accuracy and F1-scores due to clearer separation between classes. LSTMs generally perform better than Transformers across tasks, likely because of their ability to track temporal patterns in smaller, noisier datasets. Although the study uses a single dataset, testing across multiple programming tasks suggests the method can adapt to different problem types. These results support the use of gaze data and sequence models for identifying learner skill levels and informing personalized programming instruction.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"14 ","pages":"22304-22317"},"PeriodicalIF":3.6,"publicationDate":"2026-01-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11366679","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146175625","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A New Touchscreen Cover for Braille-Based Data Entry on Mobile Devices","authors":"Ghassan Aladool;Olivier Togni","doi":"10.1109/ACCESS.2026.3658020","DOIUrl":"https://doi.org/10.1109/ACCESS.2026.3658020","url":null,"abstract":"The emerging advances in mobile technology, along with the anticipated global increase in the blind and low-vision population, emphasize the importance of introducing this community to such technology as an effective and low-cost communication tool. Aiming to enable blind and deaf-blind individuals to use mobile devices, many tactile-based methods and systems have been presented in literature over the last decade. Although these methods are well-regarded, their limitations include the need for object localization on mobile devices’ embedded touchscreens and the requirement to apply multiple gesture types, often involving several fingers from both hands. Addressing these limitations, this work presents a framework offering a new tactile-based communication method for blind individuals on mobile devices, with a touchscreen cover as a core component. The invented cover splits the touchscreen into eight equally-sized cells—six for data entry and two for control input—and provides an effective solution to localization and navigation challenges faced by blind individuals on touchscreens. In particular, the proposed framework enables Braille-based character entry. An Android app has also been developed to identify entered characters and save them for further processing. The proposed framework was tested at a foundation providing care for blind and low-vision individuals. Two-phases experiments, preliminary and final, were designed and conducted with ten participants. Experimental analysis based on measuring data entry time and input error indicates that the proposed framework performs well. Furthermore, two participant surveys were conducted prior and post the experiments to justify the calculated performance measures, asses cognitive load, and verify the usability and learnability of the proposed framework.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"14 ","pages":"14931-14941"},"PeriodicalIF":3.6,"publicationDate":"2026-01-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11363568","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146082083","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Dynamic Radar Cross-Section Estimation of Chaff Clouds Based on a Surrogate Model for Spatiotemporal Distribution","authors":"Jun-Seon Kim;Uk Jin Jung;Su Hong Park;Donghyun Kim;Moonhong Kim;Dongwoo Sohn;Dong-Wook Seo","doi":"10.1109/ACCESS.2026.3657414","DOIUrl":"https://doi.org/10.1109/ACCESS.2026.3657414","url":null,"abstract":"This paper presents a novel surrogate modeling approach for estimating the dynamic radar cross-section (RCS) of chaff clouds under diverse launch and environmental conditions. A high-fidelity computational fluid dynamic–discrete element method (CFD-DEM) framework is first used to simulate the multiphysics behavior of chaff clouds generated by both naval and aircraft dispensers. These simulations generate detailed aerodynamic datasets, which are used to train a Gaussian process regression (GPR)–based surrogate model. The surrogate model enables efficient prediction of the spatiotemporal distribution of chaff clouds, incorporating variables such as wind speed, wind direction, and launch parameters. To estimate dynamic RCS, the spatiotemporal distributions are combined with approximation techniques, specifically the generalized equivalent conductor (GEC) and vector radiative transfer (VRT) methods. A real-time chaff cloud simulator with a graphical user interface is also developed, integrating aerodynamic modeling, RCS calculations, and signal fluctuation modeling. Simulation results demonstrate that the proposed surrogate model achieves high prediction accuracy, with normalized mean absolute errors (NMAE) of 0.0085 for naval chaff and 0.0176 for aircraft chaff. The dynamic RCS obtained via the surrogate model closely matches the CFD-DEM results while substantially reducing computational cost, thus offering practical utility for real-time system applications.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"14 ","pages":"14857-14869"},"PeriodicalIF":3.6,"publicationDate":"2026-01-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11363218","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146082002","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"CoastXplain: Explainable, Unsupervised Time-Series Modeling of Coastline Changes","authors":"Chandrabali Karmakar;Arnab Bhowmik;Andres Camero;Mihai Datcu","doi":"10.1109/ACCESS.2026.3656095","DOIUrl":"https://doi.org/10.1109/ACCESS.2026.3656095","url":null,"abstract":"Due to changes in global climate, coastal erosion is a common phenomenon, leading to flooding, habitat loss, property damage, and economic impacts. Large-scale automatic monitoring of coastlines is feasible by processing satellite images with Artificial Intelligence (AI) models; however, it is often constrained by scarce labels and model opacity. We present an explainable, fully unsupervised framework that first converts multi-year Sentinel-2 image series into pixel-wise land–water maps with attached model certainty and then discovers change patterns in the temporal structures of these land-water masks over a specified study period. Throughout the pipeline, we used the Normalized Difference Water Index (NDWI), computed from Sentinel-2 Green (B3) and NIR (B8), as our primary spectral indicator of surface water. We leverage two probabilistic models independently to select the best: (i) a Latent Dirichlet Allocation (LDA) model on a bag-of-visual-words to discover interpretable topics (e.g., foam, water, vegetation) that provide semantic segmentation with transparency, and (ii) a Gaussian Mixture Model (GMM) in pixel feature space to produce per-pixel probabilistic clustering and segmentation. Both models create multiple binary land-water segmentation maps from static images and then a single time-series evolution map from these multi-year land-water segmentation maps. The shoreline was automatically extracted as the boundary land pixels in binary maps with a threshold-based certainty score. We summarize changes with three methods and a score to define the amount of changes: the scene-level water fraction <inline-formula> <tex-math>$widehat {p}_{mathrm {w}}$ </tex-math></inline-formula>, shoreline-normal displacement (SND) along fixed transects, and the low-confidence proportion <inline-formula> <tex-math>${mathcal {U}}_{tau }$ </tex-math></inline-formula> that localizes ambiguity to physically dynamic interfaces. A Hamming score was computed to provide a quick and quantitative overview of the time-series evolution over a 7-year period. The pipeline delivers clear coastlines, interpretable segmentation maps, and uncertainty visualizations to help domain experts in auditing decisions. The result is a label-free, scalable, and explainable workflow in which every decision is accompanied by a confidence score, supporting reliable coastal monitoring and downstream scientific and application-oriented use.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"14 ","pages":"29628-29647"},"PeriodicalIF":3.6,"publicationDate":"2026-01-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11359192","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147299565","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Sensorless Collaborative Impedance Control of Rehabilitation Robots via LQR and Model-Based Sliding Manifold","authors":"Brahim Brahmi","doi":"10.1109/ACCESS.2026.3655024","DOIUrl":"https://doi.org/10.1109/ACCESS.2026.3655024","url":null,"abstract":"Human–robot interaction (HRI) remains a critical research area, particularly in assistive robotics, where intuitive, safe, and adaptive collaboration with human users is essential for real-world deployment. Despite significant advances in impedance-based and adaptive control strategies, many existing approaches rely on full-state measurements, exhibit limited capability in handling user intent, and lack adaptability to unstructured environments. This paper addresses these challenges by proposing a novel sensorless control architecture that integrates an optimal Linear Quadratic Regulator (LQR), a nonlinear force observer, and a robust sliding mode controller with nonlinear model-based switching function. The primary objective is to achieve accurate trajectory tracking while minimizing user-applied interaction forces without requiring force or velocity sensors. The proposed controller dynamically generates reference trajectories through a human-cooperative LQR paradigm that penalizes both robot effort and human torque, enabling adaptive behavior based on inferred user intent. A nonlinear observer estimates interaction torques using only joint position measurements, facilitating intent inference and real-time impedance adaptation. These estimates are subsequently incorporated into the construction of model-based switching manifolds within the sliding mode controller, enhancing chattering mitigation and control decoupling under uncertainty and partial state observability. Theoretical analysis establishes global asymptotic stability, and comprehensive simulations conducted on a 2-DOF rehabilitation robot validate the proposed approach under multiple disturbance scenarios. Compared to conventional impedance control strategies, the proposed method demonstrates improved accuracy, robustness, and energy efficiency, highlighting its potential for deployment in sensor-limited, human-in-the-loop applications such as prosthetics, exoskeletons, and adaptive rehabilitation robotics.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"14 ","pages":"10760-10781"},"PeriodicalIF":3.6,"publicationDate":"2026-01-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11357932","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146026330","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"PF-DTN: Predictive Routing for Intelligent Delay Tolerant Networks Using RNN-LSTM Deep Learning With Monte Carlo Dropout Uncertainty Estimation and Hybrid Deterministic, Probabilistic, and Uncertain Routing Strategies","authors":"El Mastapha Sammou","doi":"10.1109/ACCESS.2026.3655507","DOIUrl":"https://doi.org/10.1109/ACCESS.2026.3655507","url":null,"abstract":"Delay tolerant networks (DTN/s) represent an evolution of traditional ad hoc networks, specifically designed for extreme environments characterized by intermittent connectivity, unpredictable node mobility, high transmission delays, and frequent disruptions. Conventional routing protocols prove ineffective in these contexts, necessitating adaptive and robust solutions. In this paper, we propose PF-DTN (Predictive Forwarding for DTN), a hybrid adaptive routing algorithm that combines the prediction of future trajectories of DTN nodes using an LSTM (Long Short-Term Memory) model with relay selection leveraging deterministic, probabilistic, and uncertain strategies. PF-DTN also integrates uncertainty estimation via the Monte Carlo Dropout technique, enabling the dynamic adaptation of the routing strategy based on prediction reliability. The proposed PF-DTN architecture is structured into three phases: contextual mobility data collection, trajectory prediction with uncertainty estimation, and adaptive relay selection. Experimental evaluations conducted on the ONE simulator demonstrate that PF-DTN outperforms the benchmark protocol Prophet and remains competitive with recent state-of-the-art approaches. Our approach achieves a high delivery rate with relative gains over Prophet reaching up to + 46.15% in low- to medium-density networks, along with latency reduction of up to + 22.89% in high-density environments. The elevated overhead ratio, attributable to the computational demands of predictive and adaptive modules, represents a justified trade-off given the improvements in delivery reliability and latency control. These results demonstrate PF-DTN’s effectiveness and robustness across diverse DTN environments, ranging from predictable to highly dynamic and uncertain networks, establishing an optimal balance between reliability, speed, and communication cost.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"14 ","pages":"10841-10859"},"PeriodicalIF":3.6,"publicationDate":"2026-01-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11357868","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146026415","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"An Improved Template Inversion Attack Against Korean Face Images","authors":"Koy Motita;Sang Hoon Han;Younho Lee","doi":"10.1109/ACCESS.2026.3655482","DOIUrl":"https://doi.org/10.1109/ACCESS.2026.3655482","url":null,"abstract":"Recent template inversion methods achieve strong performance on Western datasets, but poor results on Korean facial images. This gap poses security risks for the authentication system using facial recognition in Korea, as the stakeholders may misunderstand that protecting facial templates on their databases is not necessary. We propose an enhanced template inversion method with three improvements: refined preprocessing using per-eye alignment and GFPGAN upscaling, MSE-enhanced loss function, and dynamic weight adjustment. Our method outperforms existing approaches, achieving 0.882 cosine similarity, <inline-formula> <tex-math>$9.7~L_{2}$ </tex-math></inline-formula> norm, and 0.312 LPIPS. Beyond reconstruction quality, we evaluate the security implications of template inversion attacks by analyzing verification robustness using Successful Attack Rate (SAR) at fixed False Match Rate (FMR) operating points, Receiver Operating Characteristic (ROC) curves, and Equal Error Rate (EER), providing a comprehensive assessment under realistic authentication thresholds. These results emphasize the critical need for robust facial template protection in facial recognition authentication systems.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"14 ","pages":"10871-10882"},"PeriodicalIF":3.6,"publicationDate":"2026-01-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11357908","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146026290","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Deep Learning-Based Faulty Component Diagnosis of Transmission Channels in ATE Affected by Thermal Degradation","authors":"Jimin Gu;Jeonghyeon Choi;Youbean Kim","doi":"10.1109/ACCESS.2026.3655081","DOIUrl":"https://doi.org/10.1109/ACCESS.2026.3655081","url":null,"abstract":"As the operating frequency of automated test equipment (ATE) increases, the thermal degradation of the components that constitute the channel accelerates. Degraded components cause signal integrity (SI) issues in the channel, which is a major factor in reducing the test quality and, thus, degrading the reliability of the ATE. Traditionally, test engineers have detected degraded components through direct probing; however, this process is time-consuming and necessitates an automated faulty component diagnosis framework. Accordingly, in this study, we propose a deep learning-based faulty component diagnosis framework to identify components that cause signal quality degradation due to heat in the ATE transmission channel. To analyze the effect of the thermal degradation of individual components on signal quality, a component modeling approach utilizing electromagnetic (EM) simulation was employed to construct a database of S-parameter data based on the temperature of the component. The simulation model demonstrated a high correlation with the measurement waveform data, with an average consistency of 97.1%, thereby ensuring its reliability. Furthermore, to address the issue of data scarcity in industrial environments, a conditional generative adversarial network (CGAN) was developed to generate S-parameter image data. The generated data showed a high similarity to the original S-parameter image data, with an average structural similarity index measure (SSIM) of 0.9845 and a peak signal-to-noise ratio (PSNR) of 35.21 dB. The convolutional neural network (CNN)-based faulty component diagnosis model trained with augmented data exhibited excellent performance, classifying faulty component types with an accuracy of 99.78%.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"14 ","pages":"11019-11034"},"PeriodicalIF":3.6,"publicationDate":"2026-01-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11357867","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146026292","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A Robust Occlusion-Aware Deep Learning Architecture for Thermal Aerial Person Classification in Search-and-Rescue Missions","authors":"Christian Woesle;Leopold Fischer-Brandies;Ricardo Buettner","doi":"10.1109/ACCESS.2026.3655411","DOIUrl":"https://doi.org/10.1109/ACCESS.2026.3655411","url":null,"abstract":"Uncrewed aerial vehicles equipped with long-wave infrared cameras are a promising tool for locating missing persons, yet their performance collapses when more than 70% of a human body is obscured by vegetation. We address this limitation by reformulating the task as a thermal classification problem rather than a detection problem, allowing the system to focus on thermal appearance cues that remain visible under heavy occlusion. Recognition robustness is further affected by high-frequency infrared sensor noise, which increases with flight altitude. Using five-fold cross-validation at flight altitudes of 30 m, 50 m, and 70 m, we show that the proposed classification pipeline achieves an accuracy of 99.07% and maintains strong recall under heavy occlusion while operating at a low computational cost. Analysis of altitude-specific models demonstrates that the Gaussian-enhanced pipeline provides a statistically significant improvement in robustness, with the largest gains observed at higher flight altitudes due to more effective suppression of altitude-dependent sensor noise. These findings establish an operational robustness benchmark for occlusion-aware thermal person classification and provide a reproducible foundation for improving the reliability of uncrewed aerial vehicle search-and-rescue systems.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"14 ","pages":"10923-10938"},"PeriodicalIF":3.6,"publicationDate":"2026-01-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11357877","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146026294","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}