Pub Date : 2026-02-09DOI: 10.1109/ACCESS.2026.3662520
Süleyman Nazmı Diker;C. Okan Sakar
Key phrase extraction (KPE) is the identification of informative phrases to provide a summary of a given document. Although unsupervised methods are highly valued for their domain independence and generalization capabilities, research in this area has predominantly focused on English texts. Conversely, KPE for agglutinative languages like Turkish remains understudied, primarily due to the challenges posed by complex morphology and a scarcity of high-quality, organized benchmarks. This study explores a systematic approach to adapting existing embedding-based KPE models, which have proven successful in English, for the Turkish language. A heuristic-based extraction method is investigated for constructing candidate pools, with attention to the impact of chunking strategies. The analysis includes three embedding-based models (AttentionRank, SAMRank, and JointGL) evaluated on two newly compiled Turkish datasets from the Information and Health Sciences domains. We also examine how well these architectures adapt and discuss their relevance for future research. While statistical methods remain strong baselines, the experimental results show that Turkish pre-trained language models (PLMs) enable existing embedding-based models to be adapted to Turkish and support the establishment of new baselines. Among these approaches, AttentionRank shows the highest performance and stability. In addition, hybrid methods that combine embedding-based and statistical techniques achieve strong performance, adapt well to different dataset properties, and provide robust baselines for future studies.
{"title":"A Systematic Approach to Key Phrase Extraction for Turkish: Adapting Embedding-Based Models","authors":"Süleyman Nazmı Diker;C. Okan Sakar","doi":"10.1109/ACCESS.2026.3662520","DOIUrl":"https://doi.org/10.1109/ACCESS.2026.3662520","url":null,"abstract":"Key phrase extraction (KPE) is the identification of informative phrases to provide a summary of a given document. Although unsupervised methods are highly valued for their domain independence and generalization capabilities, research in this area has predominantly focused on English texts. Conversely, KPE for agglutinative languages like Turkish remains understudied, primarily due to the challenges posed by complex morphology and a scarcity of high-quality, organized benchmarks. This study explores a systematic approach to adapting existing embedding-based KPE models, which have proven successful in English, for the Turkish language. A heuristic-based extraction method is investigated for constructing candidate pools, with attention to the impact of chunking strategies. The analysis includes three embedding-based models (AttentionRank, SAMRank, and JointGL) evaluated on two newly compiled Turkish datasets from the Information and Health Sciences domains. We also examine how well these architectures adapt and discuss their relevance for future research. While statistical methods remain strong baselines, the experimental results show that Turkish pre-trained language models (PLMs) enable existing embedding-based models to be adapted to Turkish and support the establishment of new baselines. Among these approaches, AttentionRank shows the highest performance and stability. In addition, hybrid methods that combine embedding-based and statistical techniques achieve strong performance, adapt well to different dataset properties, and provide robust baselines for future studies.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"14 ","pages":"22211-22231"},"PeriodicalIF":3.6,"publicationDate":"2026-02-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11381437","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146175661","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}
This study introduces a compact monopole antenna designed for near-field breast tumor detection within the 915 MHz ISM band. By integrating rectangular slots along the antenna’s non-radiating edges, the design achieves a 59% size reduction, resulting in a compact footprint of $0.4lambda _{g} times 0.27lambda _{g}$ . To enhance its performance, a partially reflective surface (PRS) is added beneath the antenna, significantly boosting gain from 1.86 dBi to 5.3 dBi, while maintaining high radiation efficiency between 95% and 99% across the 907–920 MHz range. A key feature of this study was its S-parameter analysis using a realistic three-dimensional breast phantom composed of both adipose and fibroglandular tissues. This study shows that the antenna’s reflection coefficients are sensitive to tumor size, enabling the detection of structural variations in breast tissue. Furthermore, the electromagnetic behaviour of the antenna was modelled to assess the how power is transmitted and reflected through different breast layers, with and without the PRS. Specific absorption rate (SAR) analysis confirmed that the sensor operates safely at power levels up to 215 mW. The experimental results closely aligned with simulations, confirming the antenna’s reliability and effectiveness in detecting early-stage breast cancer by identifying tumors of varying sizes in both adipose and fibroglandular tissue environments.
{"title":"Integration of a Slotted Monopole Antenna With a Partially Reflective Surface for Enhanced Breast Tumor Detection","authors":"Tiruganesh Lanka;Divya Chaturvedi;Maddirala Vijaya Lakshmi Bhavani;Arvind Kumar;Sreenivasulu Tupakula","doi":"10.1109/ACCESS.2026.3662507","DOIUrl":"https://doi.org/10.1109/ACCESS.2026.3662507","url":null,"abstract":"This study introduces a compact monopole antenna designed for near-field breast tumor detection within the 915 MHz ISM band. By integrating rectangular slots along the antenna’s non-radiating edges, the design achieves a 59% size reduction, resulting in a compact footprint of <inline-formula> <tex-math>$0.4lambda _{g} times 0.27lambda _{g}$ </tex-math></inline-formula>. To enhance its performance, a partially reflective surface (PRS) is added beneath the antenna, significantly boosting gain from 1.86 dBi to 5.3 dBi, while maintaining high radiation efficiency between 95% and 99% across the 907–920 MHz range. A key feature of this study was its S-parameter analysis using a realistic three-dimensional breast phantom composed of both adipose and fibroglandular tissues. This study shows that the antenna’s reflection coefficients are sensitive to tumor size, enabling the detection of structural variations in breast tissue. Furthermore, the electromagnetic behaviour of the antenna was modelled to assess the how power is transmitted and reflected through different breast layers, with and without the PRS. Specific absorption rate (SAR) analysis confirmed that the sensor operates safely at power levels up to 215 mW. The experimental results closely aligned with simulations, confirming the antenna’s reliability and effectiveness in detecting early-stage breast cancer by identifying tumors of varying sizes in both adipose and fibroglandular tissue environments.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"14 ","pages":"22503-22513"},"PeriodicalIF":3.6,"publicationDate":"2026-02-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11381428","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146175590","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}
Pub Date : 2026-02-09DOI: 10.1109/ACCESS.2026.3662397
Pratiwi Millati;Yoon-Ho Choi
The Automatic Identification System (AIS) generates massive volumes of real-time vessel data across vast maritime regions, making manual anomaly detection impractical for Vessel Traffic Controllers. Existing methods struggle with scalability across diverse maritime traffic and often fail to generalize to unseen vessels in dynamic environments. This paper proposes a novel two-level grid-based data representation for AIS, incorporating: 1) a discretization-based location encoder that maps vessel positions to spatial cells, 2) navigational feature weighting combining speed and course with learned importance, and 3) hierarchical anomaly localization using coarse-grained detection and fine-grained pinpointing. Our feature engineering approach attains a precision of 0.91, a recall of 0.92, an F1-score of 0.90, and an accuracy of 0.92, being tested using an unsupervised Isolation Forest model. Evaluated on the Hawaii and Bornholm Island AIS datasets, our method achieves an Area Under the Curve (AUC) of 0.75 for anomaly localization. These results demonstrate the effectiveness of our grid-based representation and feature engineering for AIS anomaly detection.
{"title":"Abnormal Vessel Activity Detection Using a Two-Level Grid Representation of AIS Data","authors":"Pratiwi Millati;Yoon-Ho Choi","doi":"10.1109/ACCESS.2026.3662397","DOIUrl":"https://doi.org/10.1109/ACCESS.2026.3662397","url":null,"abstract":"The Automatic Identification System (AIS) generates massive volumes of real-time vessel data across vast maritime regions, making manual anomaly detection impractical for Vessel Traffic Controllers. Existing methods struggle with scalability across diverse maritime traffic and often fail to generalize to unseen vessels in dynamic environments. This paper proposes a novel two-level grid-based data representation for AIS, incorporating: 1) a discretization-based location encoder that maps vessel positions to spatial cells, 2) navigational feature weighting combining speed and course with learned importance, and 3) hierarchical anomaly localization using coarse-grained detection and fine-grained pinpointing. Our feature engineering approach attains a precision of 0.91, a recall of 0.92, an F1-score of 0.90, and an accuracy of 0.92, being tested using an unsupervised Isolation Forest model. Evaluated on the Hawaii and Bornholm Island AIS datasets, our method achieves an Area Under the Curve (AUC) of 0.75 for anomaly localization. These results demonstrate the effectiveness of our grid-based representation and feature engineering for AIS anomaly detection.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"14 ","pages":"22289-22303"},"PeriodicalIF":3.6,"publicationDate":"2026-02-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11381432","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146175631","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}
Pub Date : 2026-02-09DOI: 10.1109/ACCESS.2026.3662382
Jewoo Go;Wonjin Sung
Building on the standardization efforts in 5G-Advanced for artificial intelligence (AI) and machine learning (ML)-based beam management (BM), two primary types of beam prediction models have been considered for deployment: the user equipment (UE)-sided model which is trained and deployed on the UE, and the network (NW)-sided model which is trained and deployed at the base station (BS). However, the NW-sided model suffers from significant communication overhead, while the UE-sided model is constrained by its reliance on training with data with limited diversity. To address these challenges, this paper proposes a federated learning (FL)-based approach as an alternative to centralized deployment at either the BS or the UE. The proposed FL-based model utilizes locally collected physical layer reference signal received power (L1-RSRP) data from each UE to train local models, with only the model parameters transmitted to the BS for aggregation. This approach achieves beam prediction performance comparable to that of the NW-sided model while reducing communication overhead by a factor of 4.3 and outperforming the UE-sided model by 11.2%.
{"title":"Federated Learning for Beam Management in 5G and Beyond: A Collaborative AI/ML Approach","authors":"Jewoo Go;Wonjin Sung","doi":"10.1109/ACCESS.2026.3662382","DOIUrl":"https://doi.org/10.1109/ACCESS.2026.3662382","url":null,"abstract":"Building on the standardization efforts in 5G-Advanced for artificial intelligence (AI) and machine learning (ML)-based beam management (BM), two primary types of beam prediction models have been considered for deployment: the user equipment (UE)-sided model which is trained and deployed on the UE, and the network (NW)-sided model which is trained and deployed at the base station (BS). However, the NW-sided model suffers from significant communication overhead, while the UE-sided model is constrained by its reliance on training with data with limited diversity. To address these challenges, this paper proposes a federated learning (FL)-based approach as an alternative to centralized deployment at either the BS or the UE. The proposed FL-based model utilizes locally collected physical layer reference signal received power (L1-RSRP) data from each UE to train local models, with only the model parameters transmitted to the BS for aggregation. This approach achieves beam prediction performance comparable to that of the NW-sided model while reducing communication overhead by a factor of 4.3 and outperforming the UE-sided model by 11.2%.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"14 ","pages":"22434-22444"},"PeriodicalIF":3.6,"publicationDate":"2026-02-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11381446","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146175632","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}
Pub Date : 2026-02-09DOI: 10.1109/ACCESS.2026.3662521
Jean Carlos Bortoli Dalcin;Marlio José Do Couto Bonfim
The advancement of modern technologies and the growing demand for high-speed communication systems that integrate digital and analog circuits into multilayer printed circuit boards (PCBs) require the development of efficient solutions to ensure signal integrity (SI), power integrity (PI), and electromagnetic interference mitigation. In this context, this article proposes the design and implementation of a new Electromagnetic Band Gap (EBG) structure, referred to as MS-EBG, with multiple applications, including Common-Mode (CM) noise suppression, Simultaneous Switching Noise (SSN) mitigation, and power integrity preservation. The proposed MS-EBG structure was developed with compact dimensions of $20 times 20$ mm and exhibited attenuation levels of up to 40 dB over the frequency range from 1.82 to 18 GHz. The design methodology included equivalent circuit modeling, the incorporation of Defected Ground Structures (DGS), electromagnetic simulations, and experimental validation. The results demonstrate that the MS-EBG structure achieves superior SSN rejection compared to previous studies, while also significantly reducing near-field radiation above 2 GHz. Power integrity measurements revealed an increase in impedance relative to the reference board, although the values remained within acceptable limits for practical applications. Therefore, the MS-EBG structure is shown to be an effective solution for mitigating CM noise and SSN over a wide frequency range, offering reduced physical dimensions compared to alternatives reported in the literature, while simultaneously maintaining both signal and power integrity. Consequently, the proposed structure presents itself as a viable and efficient solution for various applications across broad operational frequency ranges.
{"title":"Hybrid EBG Structure for Common-Mode Noise and SSN Suppression in Multilayer PCBs","authors":"Jean Carlos Bortoli Dalcin;Marlio José Do Couto Bonfim","doi":"10.1109/ACCESS.2026.3662521","DOIUrl":"https://doi.org/10.1109/ACCESS.2026.3662521","url":null,"abstract":"The advancement of modern technologies and the growing demand for high-speed communication systems that integrate digital and analog circuits into multilayer printed circuit boards (PCBs) require the development of efficient solutions to ensure signal integrity (SI), power integrity (PI), and electromagnetic interference mitigation. In this context, this article proposes the design and implementation of a new Electromagnetic Band Gap (EBG) structure, referred to as MS-EBG, with multiple applications, including Common-Mode (CM) noise suppression, Simultaneous Switching Noise (SSN) mitigation, and power integrity preservation. The proposed MS-EBG structure was developed with compact dimensions of <inline-formula> <tex-math>$20 times 20$ </tex-math></inline-formula> mm and exhibited attenuation levels of up to 40 dB over the frequency range from 1.82 to 18 GHz. The design methodology included equivalent circuit modeling, the incorporation of Defected Ground Structures (DGS), electromagnetic simulations, and experimental validation. The results demonstrate that the MS-EBG structure achieves superior SSN rejection compared to previous studies, while also significantly reducing near-field radiation above 2 GHz. Power integrity measurements revealed an increase in impedance relative to the reference board, although the values remained within acceptable limits for practical applications. Therefore, the MS-EBG structure is shown to be an effective solution for mitigating CM noise and SSN over a wide frequency range, offering reduced physical dimensions compared to alternatives reported in the literature, while simultaneously maintaining both signal and power integrity. Consequently, the proposed structure presents itself as a viable and efficient solution for various applications across broad operational frequency ranges.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"14 ","pages":"22491-22502"},"PeriodicalIF":3.6,"publicationDate":"2026-02-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11381579","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146175660","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}
Pub Date : 2026-02-09DOI: 10.1109/ACCESS.2026.3662344
Ayesha Khaliq;Kolawole J. Adebayo
Large Language Models (LLMs) often face severe latency and computational cost constraints which hinder their adoption in real-time enterprise applications. Retrieval-Augmented Generation (RAG) frameworks, while improving factual accuracy, further increase inference delays owing to additional retrieval and context integration steps. To address these challenges, we propose Dynamic Knowledge Caching for Large Language Models (DKC-LLM), a novel framework that integrates dynamic semantic caching with an adaptive cache management strategy, which detects and compensates for semantic drift to accelerate response generation while ensuring accuracy and freshness. We evaluate DKC-LLM on two distinct question-answering benchmarks: BankFAQs, a domain-specific dataset with 5,000 selected queries and HotpotQA, an open-domain dataset with 5,000 selected queries, totaling 10,000 evaluation queries. Experimental results demonstrate that DKC-LLM achieves 20-30% lower latency, i.e., 5.3–6.8 ms vs. RAG’s 8–10 ms, a 63-70.5% cache hit rate, and a 60% reduction in LLM compute overhead, while maintaining 83-90% response accuracy. Additionally, DKC-LLM reduced hallucination rates by 75%, outperforming baseline RAG by up to 15 percentage points. These findings posit that DKC-LLM is a cost-efficient, low-latency, and high-accuracy solution for real-time, high-frequency business scenarios such as customer support and enterprise information retrieval.
{"title":"DKC-LLM: Dynamic Knowledge Caching for Large Language Models in Business Applications","authors":"Ayesha Khaliq;Kolawole J. Adebayo","doi":"10.1109/ACCESS.2026.3662344","DOIUrl":"https://doi.org/10.1109/ACCESS.2026.3662344","url":null,"abstract":"Large Language Models (LLMs) often face severe latency and computational cost constraints which hinder their adoption in real-time enterprise applications. Retrieval-Augmented Generation (RAG) frameworks, while improving factual accuracy, further increase inference delays owing to additional retrieval and context integration steps. To address these challenges, we propose Dynamic Knowledge Caching for Large Language Models (DKC-LLM), a novel framework that integrates dynamic semantic caching with an adaptive cache management strategy, which detects and compensates for semantic drift to accelerate response generation while ensuring accuracy and freshness. We evaluate DKC-LLM on two distinct question-answering benchmarks: BankFAQs, a domain-specific dataset with 5,000 selected queries and HotpotQA, an open-domain dataset with 5,000 selected queries, totaling 10,000 evaluation queries. Experimental results demonstrate that DKC-LLM achieves 20-30% lower latency, i.e., 5.3–6.8 ms vs. RAG’s 8–10 ms, a 63-70.5% cache hit rate, and a 60% reduction in LLM compute overhead, while maintaining 83-90% response accuracy. Additionally, DKC-LLM reduced hallucination rates by 75%, outperforming baseline RAG by up to 15 percentage points. These findings posit that DKC-LLM is a cost-efficient, low-latency, and high-accuracy solution for real-time, high-frequency business scenarios such as customer support and enterprise information retrieval.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"14 ","pages":"22318-22334"},"PeriodicalIF":3.6,"publicationDate":"2026-02-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11380191","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146175623","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}
Pub Date : 2026-02-09DOI: 10.1109/ACCESS.2026.3663058
Raghav Krishna;Dilip Kumar Choudhary
Chaotic dynamical systems, such as the Lorenz, Rössler, and Duffing attractors, are studied for their irregular trajectories and broadband spectra. However, a unified framework for their fingerprinting and quantitative measures is not widely present. Existing methods focus on either spectral methods or dynamical complexity measures in isolation, which limits interpretability and robustness across diverse regimes. In this work, we propose an entropy based spectral fingerprinting framework that integrates the Characteristic Frequency Score (CFS) obtained from short time Fourier transform (STFT) and Continuous Wavelet Transform (CWT) with nonlinear complexity metrics including the largest Lyapunov exponent (LLE), approximate entropy (ApEn), and permutation entropy (PermEn). Using clean signal analysis, we show that the Duffing oscillator produces compact spectral concentration with high instability and low entropy, the Lorenz system exhibits broadband spreading with higher entropic complexity, and the Rössler attractor displays intermediate behavior. Confidence intervals which are derived through bootstrap resampling confirm the statistical stability of the extracted measures, with STFT emphasizing global spectral spread and CWT highlighting localized structures. The proposed framework offers an interpretable and statistically consistent means of fingerprinting chaotic systems, with applications in secure communication, biomedical signal monitoring, fault diagnostics, and complexity science.
{"title":"Unified Entropy–Spectral Fingerprinting of Chaotic Attractors via CFS, Lyapunov Stability, and Nonlinear Complexity Measures","authors":"Raghav Krishna;Dilip Kumar Choudhary","doi":"10.1109/ACCESS.2026.3663058","DOIUrl":"https://doi.org/10.1109/ACCESS.2026.3663058","url":null,"abstract":"Chaotic dynamical systems, such as the Lorenz, Rössler, and Duffing attractors, are studied for their irregular trajectories and broadband spectra. However, a unified framework for their fingerprinting and quantitative measures is not widely present. Existing methods focus on either spectral methods or dynamical complexity measures in isolation, which limits interpretability and robustness across diverse regimes. In this work, we propose an entropy based spectral fingerprinting framework that integrates the Characteristic Frequency Score (CFS) obtained from short time Fourier transform (STFT) and Continuous Wavelet Transform (CWT) with nonlinear complexity metrics including the largest Lyapunov exponent (LLE), approximate entropy (ApEn), and permutation entropy (PermEn). Using clean signal analysis, we show that the Duffing oscillator produces compact spectral concentration with high instability and low entropy, the Lorenz system exhibits broadband spreading with higher entropic complexity, and the Rössler attractor displays intermediate behavior. Confidence intervals which are derived through bootstrap resampling confirm the statistical stability of the extracted measures, with STFT emphasizing global spectral spread and CWT highlighting localized structures. The proposed framework offers an interpretable and statistically consistent means of fingerprinting chaotic systems, with applications in secure communication, biomedical signal monitoring, fault diagnostics, and complexity science.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"14 ","pages":"22390-22403"},"PeriodicalIF":3.6,"publicationDate":"2026-02-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11386824","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146175630","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}
Modular Multilevel Converters (MMCs) are widely adopted in high-power applications such as renewable energy integration and HVDC transmission, owing to their modularity and efficiency. Nevertheless, their complex architecture makes them susceptible to critical failures, particularly IGBT short-circuit faults, which can severely compromise system reliability. This paper presents a hybrid fault diagnosis approach that combines Discrete Wavelet Transform (DWT) and Radial Basis Function Neural Networks (RBFNN) for online detection and localization of IGBT short-circuit faults in MMCs. Capacitor voltage signals are decomposed using DWT, and the Root Mean Square (RMS) values of detail coefficients are extracted as features. These features are then processed by an RBFNN to accurately identify fault type and location. Simulation results demonstrate that the proposed method achieves fast, reliable, and accurate fault diagnosis, thereby enhancing the operational safety and stability of MMC-based systems.
{"title":"Online IGBT Short-Circuit Fault Diagnosis in Modular Multilevel Converter Based on DWT and RBFNN","authors":"Kheireddine Moulay;Imad Merzouk;Khaled Omer Mokhtar Touati;Ahmed Hafaifa;Ghada Alhussein","doi":"10.1109/ACCESS.2026.3662514","DOIUrl":"https://doi.org/10.1109/ACCESS.2026.3662514","url":null,"abstract":"Modular Multilevel Converters (MMCs) are widely adopted in high-power applications such as renewable energy integration and HVDC transmission, owing to their modularity and efficiency. Nevertheless, their complex architecture makes them susceptible to critical failures, particularly IGBT short-circuit faults, which can severely compromise system reliability. This paper presents a hybrid fault diagnosis approach that combines Discrete Wavelet Transform (DWT) and Radial Basis Function Neural Networks (RBFNN) for online detection and localization of IGBT short-circuit faults in MMCs. Capacitor voltage signals are decomposed using DWT, and the Root Mean Square (RMS) values of detail coefficients are extracted as features. These features are then processed by an RBFNN to accurately identify fault type and location. Simulation results demonstrate that the proposed method achieves fast, reliable, and accurate fault diagnosis, thereby enhancing the operational safety and stability of MMC-based systems.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"14 ","pages":"22252-22273"},"PeriodicalIF":3.6,"publicationDate":"2026-02-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11381407","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146175662","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}
To reconcile the conflict between phase-picking accuracy and edge-deployment efficiency, this study proposes a Lightweight Wavelet Convolutional U-Net. The core originality lies in replacing standard sampling with an invertible wavelet paradigm. We introduce a Lightweight Wavelet Convolutional Attention Module that integrates Discrete Wavelet Transform with lightweight convolutions to capture multi-scale structural cues without information loss. Furthermore, a Frequency Branch Processing and Fusion Module is devised to decouple frequency bands and execute high-fidelity reconstruction via Inverse-Discrete Wavelet Transform, effectively resolving spectral aliasing issues inherent in traditional upsampling. This frequency-aware architecture establishes a new methodology for real-time seismic monitoring, delivering a robust balance of low latency and high structural fidelity on embedded platforms.
{"title":"Lightweight Wavelet Convolutional U-Net for Seismic Phase Recognition","authors":"Jianxian Cai;Jingqi Ma;Yanxiong Wu;Ping Li;Zhuangqi Guo","doi":"10.1109/ACCESS.2026.3662343","DOIUrl":"https://doi.org/10.1109/ACCESS.2026.3662343","url":null,"abstract":"To reconcile the conflict between phase-picking accuracy and edge-deployment efficiency, this study proposes a Lightweight Wavelet Convolutional U-Net. The core originality lies in replacing standard sampling with an invertible wavelet paradigm. We introduce a Lightweight Wavelet Convolutional Attention Module that integrates Discrete Wavelet Transform with lightweight convolutions to capture multi-scale structural cues without information loss. Furthermore, a Frequency Branch Processing and Fusion Module is devised to decouple frequency bands and execute high-fidelity reconstruction via Inverse-Discrete Wavelet Transform, effectively resolving spectral aliasing issues inherent in traditional upsampling. This frequency-aware architecture establishes a new methodology for real-time seismic monitoring, delivering a robust balance of low latency and high structural fidelity on embedded platforms.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"14 ","pages":"22372-22389"},"PeriodicalIF":3.6,"publicationDate":"2026-02-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11381422","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146175628","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}
Dementia is a growing global health challenge, making early detection of cognitive decline critically important. Early detection of dementia and other cognitive impairment is essential for timely intervention and better care planning. However, existing datasets for training automated screening tools are limited, especially for underrepresented languages such as Czech. In this study, we present a new dataset and a novel application named the DigiDiaDem (Digital Diagnostics of Dementia) designed for automated dementia screening through multimodal cognitive assessment. The application integrates user-friendly digital cognitive tasks with machine learning algorithms to evaluate linguistic and cognitive performance in real time. Using this system, we collected and curated a comprehensive dataset of speech and cognitive data from Czech-speaking participants. The dataset comprises 371 individuals, including cognitively normal individuals and patients with mild cognitive impairment and mild dementia. It includes socio-demographic data, results of cognitive and speech tests, functional assessment questionnaires, data collected through the DigiDiaDem application, and automatic speech recognition (ASR) transcripts of spoken responses. Raw audio recordings are not included. Instead, the dataset provides manually engineered linguistic and acoustic features. We describe the data collection process and outline the cognitive tasks used to collect the dataset. Our experiments demonstrate that speech features derived from cognitively demanding tasks, such as verbal fluency and memory recall, can effectively distinguish healthy participants from those with cognitive impairment. Models trained on the dataset achieved up to 95% accuracy when combining speech features with demographic information. Preliminary experiments demonstrate the feasibility of using the collected data for dementia detection. These findings confirm that speech-based digital assessment can complement traditional clinical evaluation. The proposed dataset and application offer a substantial resource for the research community by establishing a solid baseline for machine-learning-based approaches to dementia screening from speech-based interaction.
{"title":"The DigiDiaDem Speech-Cognitive Dataset: Initial Experiments on Detecting Cognitive Impairments From Speech","authors":"Luboš Šmídl;Filip Polák;Lucie Zajícová;Tomáš Lebeda;Jan Švec;Jan Tupý;Martin Bulín;Aleš Bartoš","doi":"10.1109/ACCESS.2026.3662045","DOIUrl":"https://doi.org/10.1109/ACCESS.2026.3662045","url":null,"abstract":"Dementia is a growing global health challenge, making early detection of cognitive decline critically important. Early detection of dementia and other cognitive impairment is essential for timely intervention and better care planning. However, existing datasets for training automated screening tools are limited, especially for underrepresented languages such as Czech. In this study, we present a new dataset and a novel application named the DigiDiaDem (Digital Diagnostics of Dementia) designed for automated dementia screening through multimodal cognitive assessment. The application integrates user-friendly digital cognitive tasks with machine learning algorithms to evaluate linguistic and cognitive performance in real time. Using this system, we collected and curated a comprehensive dataset of speech and cognitive data from Czech-speaking participants. The dataset comprises 371 individuals, including cognitively normal individuals and patients with mild cognitive impairment and mild dementia. It includes socio-demographic data, results of cognitive and speech tests, functional assessment questionnaires, data collected through the DigiDiaDem application, and automatic speech recognition (ASR) transcripts of spoken responses. Raw audio recordings are not included. Instead, the dataset provides manually engineered linguistic and acoustic features. We describe the data collection process and outline the cognitive tasks used to collect the dataset. Our experiments demonstrate that speech features derived from cognitively demanding tasks, such as verbal fluency and memory recall, can effectively distinguish healthy participants from those with cognitive impairment. Models trained on the dataset achieved up to 95% accuracy when combining speech features with demographic information. Preliminary experiments demonstrate the feasibility of using the collected data for dementia detection. These findings confirm that speech-based digital assessment can complement traditional clinical evaluation. The proposed dataset and application offer a substantial resource for the research community by establishing a solid baseline for machine-learning-based approaches to dementia screening from speech-based interaction.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"14 ","pages":"22232-22251"},"PeriodicalIF":3.6,"publicationDate":"2026-02-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11373370","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146175642","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}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：