IEEE Access最新文献

{"title":"Innovative Approaches for PCB Image Reconstruction: Tailored Datasets, Metrics, and Models","authors":"Qianyue Wang","doi":"10.1109/ACCESS.2025.3554509","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3554509","url":null,"abstract":"The reconstruction of PCB (Printed Circuit Board) images is vital for quality control in electronic manufacturing, enabling fault analysis, reverse engineering, and repair. However, this task faces significant challenges due to the high complexity and precision required, compounded by densely packed layouts and image degradation from physical damage, contamination, low-light conditions, or tampered components. Existing methods often fall short due to the lack of specialized datasets, insufficient validation against real-world degradation, limited robustness to diverse defect scenarios, and evaluation metrics that fail to capture the fine details critical for PCB functionality. To address these challenges, this study introduces a data-centric approach for PCB image reconstruction, featuring three major contributions: 1) A rigorously validated dataset, PCB_Reconstruction, comprising three subsets—PCB_Mask, PCB_Blur, and PCB_Replace—systematically simulates physical damage, contamination effects, and tampered components. The dataset’s realism is validated through Structural Similarity Index Measure (SSIM) analysis and a CNN-based classification test, confirming its alignment with real-world PCB defects. 2) A new evaluation metric, Detail Reconstruction Quality (DRQ), designed to measure edge reconstruction precision, addressing the limitations of traditional metrics like PSNR and MSE. The validity of DRQ is further demonstrated by benchmarking it against SSIM and PSNR on diverse datasets, including DIV2K and Kodak, where DRQ achieves superior fine-detail reconstruction performance. 3) Comprehensive benchmarking across seven models, providing a robust evaluation of both traditional and state-of-the-art approaches. Among these, the Non-Autoregressive Transformer (NAT) stands out for its lightweight architecture and superior edge restoration performance, achieving a 11.2% DRQ improvement over U-Net on PCB_Mask and a 9.8% DRQ increase over MAE on PCB_Blur. Retraining traditional models (e.g., U-Net and EDSR) on the PCB_Reconstruction subsets further enhances DRQ scores by an average of 15%, demonstrating the dataset’s capacity to improve robustness and generalization. Experimental results confirm that the proposed datasets, evaluation framework, and benchmarking approach significantly advance the precision and reliability of PCB image reconstruction. The datasets, evaluation metrics, and results will be made publicly available at <uri>https://github.com/Wangq180/PCB_Research</uri>.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"54267-54283"},"PeriodicalIF":3.4,"publicationDate":"2025-03-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10938583","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143748851","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":"Prefix Tuning Using Residual Reparameterization","authors":"Youngjun Jung;Hyunsun Hwang;Changki Lee","doi":"10.1109/ACCESS.2025.3554513","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3554513","url":null,"abstract":"Fine-tuning large language models for specific tasks requires updating and storing all parameters, leading to significant computational and storage cost issues. To address these challenges, parameter-efficient learning such as prefix tuning has gained attention. However, prefix tuning can suffer from sensitivity to prefix length. This is considered to be due to different prefix tokens being required for each task. In this paper, we propose improving the robustness and performance of prefix tuning through residual reparameterization. We add residual connections to the prefix module, providing more flexibility to the model. Additionally, we propose a gate mechanism to assign weights to prefix tokens, allowing for focus on more important tokens. Our experiments on the GLUE benchmark and E2E dataset demonstrate that our methods lead to improved and stabilized performance across various prefix lengths. The residual connections enable faster convergence during training, while the gate mechanism helps balance prefix tokens and find more optimized parameters. Our approach shows particular effectiveness when combining residual connections with the gate mechanism, outperforming original prefix tuning, especially with longer prefix lengths, while remaining parameter-efficient. We also provide an analysis of gate weight trends during training, offering insights into how the model uses prefix tokens for different prefix lengths.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"54866-54872"},"PeriodicalIF":3.4,"publicationDate":"2025-03-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10938609","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143761540","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 Concise Dataset for Intelligent Behaviors in Domestic Tasks","authors":"Marlon de Oliveira Vaz;Ronnier Frates Rohrich;João Alberto Fabro;André Schneider de Oliveira","doi":"10.1109/ACCESS.2025.3554607","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3554607","url":null,"abstract":"For service robots operating in domestic environments, high-level intelligent behaviors require a comprehensive understanding of objects through visual perception. The random placement of objects introduces variations that impact the accuracy of object detection and recognition. This study presents a novel method for automatically generating a concise image dataset, named the Object Dataset Federal University of Technology (ODUTF), to enhance intelligent behaviors in service robots to focus on domestic tasks. The dataset is produced using an automatic multicapture device that gathers RGB images, stereo information, depth images, and point-cloud data. This device has two degrees of freedom to adjust both the orientation of objects and the camera’s viewpoint. The method creates a precise and detailed visual description of objects, which improves a service robot’s ability to approach and pick up objects. This approach is evaluated within the context of the RoboCup@Home Brazil League, part of the international RoboCup competition dedicated to domestic service robots. This league involves diverse tasks, emphasizing object detection and recognition. The use of high-level intelligent behaviors is critical for overcoming domestic challenges, and ODUTF facilitates the deployment of more reliable deep neural network methods for tracking objects during pick-up tasks. Furthermore, ODUTF can be dynamically adapted using post-processing scripts to incorporate artificial features like varying backgrounds, luminosity, and noise.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"54722-54738"},"PeriodicalIF":3.4,"publicationDate":"2025-03-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10938585","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143761436","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":"Optimization of Railway Transportation Planning by Combining TST Model and Genetic Algorithm","authors":"Wei Cao;Fan Chen","doi":"10.1109/ACCESS.2025.3554677","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3554677","url":null,"abstract":"Railway transportation, a key long - distance freight transport method, faces challenges due to the rapid growth of global logistics demand. These challenges include high transportation costs, low punctuality rates, and inefficient resource utilization. Traditional static optimization methods cannot adapt to dynamic changes and multi-objective optimization requirements. The study proposes an integrated method that combines the Temporal-Spatial Tunnel (TST) model with the Genetic Algorithm (GA). The TST model describes railway transportation changes dynamically by integrating temporal and spatial dimensions. The GA uses its global search ability to optimize train routing and timetabling. The proposed method enhances the efficiency and flexibility of the railway transportation system. It addresses the issues of low punctuality, inefficient resource utilization, and lack of adaptability to dynamic changes and multi - objective optimization in traditional methods. Experimental results show the superiority of this approach. In urban network scenarios, it achieves a punctuality rate of 94.87%, resource utilization of 89.78%, and a response time of 280.12 seconds. In freight - priority scenarios, the maximum punctuality rate reaches 95.45%. Compared to traditional methods, it significantly improves transportation efficiency and flexibility in multi - objective optimization, offering an effective solution for railway transportation planning under dynamic demands and valuable references for logistics system scheduling optimization.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"53266-53275"},"PeriodicalIF":3.4,"publicationDate":"2025-03-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10938531","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143761299","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":"PAPR Reduction for OFDM Mobile Wireless Communications by Constant-Amplitude Signal Decomposition","authors":"Yuyuan Chang;Kazuhiko Fukawa","doi":"10.1109/ACCESS.2025.3554660","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3554660","url":null,"abstract":"This paper proposes a constant-amplitude (CA) modulation scheme for orthogonal frequency-division multiplexing (OFDM) over mobile wireless channels. To cope with the high peak-to-average power ratio (PAPR) of OFDM signals, the proposed scheme decomposes one OFDM signal into two CA signals, and then transmits these two signals at different timings. Since CA signals give off a considerable amount of out-of-band emission (OOBE), the decomposed CA signals are passed into a low-pass filter (LPF) of which outputs are sequentially transmitted from one antenna. On the receiver side, the received signals are fed into a frequency-domain equalizer and then, combining the two time-domain equalizer’s outputs that originate from those two CA signals can regenerate the original OFDM signal. Especially when the channel is time-invariant without phase noise, lower complexity recovery of the original OFDM signal is to combine and equalize the two received signals that correspond to those two CA signals. As a conventional CA OFDM approach, the constant envelope OFDM (CE-OFDM) scheme transforms the OFDM signal into a phase modulated signal and is compared with the proposed scheme by computer simulation. It is demonstrated that the proposed scheme with LPF is superior to CE-OFDM in terms of both bit error rate (BER) performance and OOBE characteristics. Computational complexity of the proposed scheme is almost the same as that of CE-OFDM without any unwrappers in the phase demodulator, but the former can require a less amount of complexity than that of the latter with an accurate unwrapper being implemented.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"54672-54685"},"PeriodicalIF":3.4,"publicationDate":"2025-03-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10938533","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143748730","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":"Interior Innovation Design Using ResNet Neural Network and Intelligent Human–Computer Interaction","authors":"Kejing Guo;Jinxin Ma","doi":"10.1109/ACCESS.2025.3554553","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3554553","url":null,"abstract":"This study aims to optimize the interior design generation process while enhancing personalization and efficiency through the development of a hybrid Residual Network-Human-Computer Interaction (ResNet-HCI) framework. It employs Residual Network (ResNet) neural networks, which leverage residual learning to improve training stability and feature extraction capabilities in deep networks. This allows for efficient feature reuse and optimization of model performance. Additionally, Human-Computer Interaction (HCI) technologies, such as voice commands, gesture control, and Virtual Reality/Augmented Reality (VR/AR), are integrated to enhance user interaction with the design system, thereby improving the intelligence and personalization of the interior design workflow. The Large-Scale Scene Understanding dataset is used for model training to evaluate system performance under varying training steps, hyperparameter configurations, and noise conditions. The experimental results show the following: 1) Significant performance variations are observed across different models under conditions such as increased training iterations, noise interference, and design scoring. In the iteration experiment, model performance generally improves with more training steps. ResNet50 consistently outperforms other models, achieving an F1 score of 0.935 after 20 iterations, demonstrating exceptional feature learning and stability. In the noise robustness analysis, ResNet50 and ResNeXt show minimal performance degradation under Gaussian noise, indicating strong noise robustness. 2) Regarding interior design scoring, hybrid layouts generated using ResNet models combined with HCI technologies excel in multiple dimensions, achieving the highest overall satisfaction scores and emerging as the optimal design solutions. These findings validate the exceptional performance and versatility of ResNet50 and ResNeXt in both deep learning and HCI applications. This study provides both theoretical and practical support for the intelligent transformation of the interior design field, while also offering insights into broader applications in creative industries.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"55130-55139"},"PeriodicalIF":3.4,"publicationDate":"2025-03-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10938593","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143761516","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":"Knowledge-Enriched Recommendations: Bridging the Gap in Alloy Material Selection With Large Language Models","authors":"Tongwei Wu;Shiyu Du;Yiming Zhang;Honggang Li","doi":"10.1109/ACCESS.2025.3554125","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3554125","url":null,"abstract":"Navigating materials performance databases efficiently is a persistent challenge in materials science and engineering, particularly in the selection of alloy materials. While recommendation systems address information overload, traditional approaches relying on historical user data face limitations such as data sparsity and cold-start issues. This study presents a novel recommendation model that integrates domain-specific knowledge graphs with large language models (LLMs) to enhance recommendation accuracy in alloy material selection. A knowledge graph for alloys is developed, encapsulating technical material data and relationships to improve retrieval and recommendation outcomes. LLMs are employed for label clustering and natural language-based instruction-following to craft user profiles and enhance data representation. Two graph enhancement strategies, integrated with attention mechanisms, effectively capture user preferences. Experimental results on a ferroalloy dataset demonstrate the model’s superior performance compared to baseline methods, significantly addressing data sparsity while offering personalized, accurate recommendations. This research bridges the gap between knowledge graphs and LLMs in recommendation systems, contributing a flexible, intelligent solution to streamline material selection processes.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"53124-53139"},"PeriodicalIF":3.4,"publicationDate":"2025-03-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10937740","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143740406","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":"Analysis of Output Characteristics of Electrically Excited Doubly Salient Generator Based on Back EMF Oriented Vector Control Strategy","authors":"Weifeng Liu;Yulong Tao;Xiaoli Meng","doi":"10.1109/ACCESS.2025.3553902","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3553902","url":null,"abstract":"Compared to using uncontrolled rectifier, the use of active rectifier can significantly improve the output power of doubly salient electric generator (DSEG). The traditional six-state-angle-control (SSAC) strategy has limited effectiveness in improving the output power of DSEG. A mathematical model of DSEG in a rotating coordinate system oriented by back electromotive force (EMF) vector was established based on the principle of instantaneous reactive power. The reason why SSAC strategy is difficult to further improve the output power of DSEG was analyzed through this model. To address this issue, this paper proposed a back EMF oriented vector control (BEFOVC) strategy. Compared with SSAC strategy, BEFOVC can further optimize the output performance of DSEG, reduce copper loss, phase current peak and THD, and suppress the torque ripple of DSEG. Finally, the feasibility and effectiveness of the proposed method were verified through simulations and experiments on a 12/8-pole DSEG.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"53254-53265"},"PeriodicalIF":3.4,"publicationDate":"2025-03-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10937377","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143761539","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":"Continual Test-Time Adaptation for Robust Remote Photoplethysmography Estimation","authors":"Hyunwoo Lee;Haechang Lee;Byung Hyun Lee;Se Young Chun","doi":"10.1109/ACCESS.2025.3554508","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3554508","url":null,"abstract":"Remote photoplethysmography (rPPG) estimation has made considerable progress by leveraging deep learning, yet its performance remains highly susceptible to the domain shifts caused by lighting, skin tone and movement, particularly during inference. Moreover, continuous adaptation across multiple domains is also challenging due to the dynamic environmental changes such as lighting transitions or continuous motions. Domain adaptation has been widely investigated, mostly focusing on classification tasks with labels. Prior arts to address domain shifts in rPPG estimation, which is a regression task, rely on labeled data, require pretraining on target domains, and focus on single-domain test-time adaptation (TTA). However, there are still remaining challenges in TTA for their applicability of rPPG estimation in real-world scenarios such as no label during inference, continuous adaptation over multiple domains, and potential catastrophic forgetting when re-adapting to the source domain. In this work, we recast an rPPG TTA problem as a continual learning and propose an efficient continual TTA method that mitigates significant domain shifts in multiple target domains without labels by leveraging the non-contrastive unsupervised learning loss with selective updates of the batch normalization layers only as well as alleviates catastrophic forgetting in source domain by adopting the learning without forgetting (LwF) regularization in the frequency domain. Our method without target labels consistently yielded improved performance in challenging continual adaptation scenarios, including adapting to multiple new domain datasets over several cycles. This approach not only mitigates catastrophic forgetting in the source domain, but also ensures robust performance across different domains.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"55049-55060"},"PeriodicalIF":3.4,"publicationDate":"2025-03-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10938539","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143748735","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":"Modality-Guided Refinement Learning for Multimodal Emotion Recognition","authors":"Sunyoung Cho","doi":"10.1109/ACCESS.2025.3554708","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3554708","url":null,"abstract":"Multimodal emotion recognition (MER) aims to understand human emotions by leveraging multiple modalities. Previous MER methods have focused on learning enhanced multimodal representations through various interaction and fusion mechanisms, utilizing different types of features from individual modalities. However, these methods often fail to account for the varying contributions of each modality to emotion, leading to suboptimal representations. To address this, we propose a modality-guided refinement learning framework that enhances multimodal representations by incorporating modality information. Specifically, we decouple multimodal representations into modality-invariant and modality-specific components by introducing shared and private encoders, which are learned by leveraging the distributional properties of the representations in their latent subspaces, guided by a modality classifier. Our method introduces margin constraints to further refine these decoupled representations, adaptively considering the contribution of each modality during the decoupling and multimodal learning processes. This optimization reduces information loss and corruption, resulting in more robust and discriminative multimodal representation learning. We evaluate our proposed method through experiments on two benchmark MER datasets: the CMU Multimodal Corpus of Sentiment Intensity (CMU-MOSI) and the CMU Multimodal Opinion Sentiment and Emotion Intensity (CMU-MOSEI). Comprehensive experiments demonstrate that our method outperforms several baseline models in multimodal emotion recognition.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"53558-53567"},"PeriodicalIF":3.4,"publicationDate":"2025-03-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10938535","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143740167","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}