Nothing Special   »   [go: up one dir, main page]

Next Issue
Volume 14, March-1
Previous Issue
Volume 14, February-1
You seem to have javascript disabled. Please note that many of the page functionalities won't work as expected without javascript enabled.
 
 

Electronics, Volume 14, Issue 4 (February-2 2025) – 181 articles

Cover Story (view full-size image): The integration of inductive charging in EVs has gained interest, especially for AUVs. In seawater, one of the effects that should be analyzed is the appearance of parasitic capacitances (Ce) between the power coils due to high conductivity. This component reduces efficiency. The main objective of this contribution is to find the optimal solution to avoid the effects of Ce during the coils design, thus simplifying the process and making it similar to an air-based solution. A comparative analysis of different topologies examines voltage, current, and efficiency variations. SS and LCC-S compensation systems are the base of this comparison. Experimental validation confirms that optimizing coil design can effectively neglect Ce effects, enhancing wireless charging efficiency for AUVs. View this paper
  • Issues are regarded as officially published after their release is announced to the table of contents alert mailing list.
  • You may sign up for e-mail alerts to receive table of contents of newly released issues.
  • PDF is the official format for papers published in both, html and pdf forms. To view the papers in pdf format, click on the "PDF Full-text" link, and use the free Adobe Reader to open them.
Order results
Result details
Section
Select all
Export citation of selected articles as:
20 pages, 436 KiB  
Article
Data-Driven Distributionally Robust Optimal Power Flow for Distribution Grids Under Wasserstein Ambiguity Sets
by Fangzhou Liu, Jincheng Huo, Fengfeng Liu, Dongliang Li and Dong Xue
Electronics 2025, 14(4), 822; https://doi.org/10.3390/electronics14040822 - 19 Feb 2025
Abstract
The increasing integration of distributed energy resources into distribution feeders introduces significant uncertainties, stemming from volatile renewable sources and other fluctuating electrical elements, which pose substantial challenges for optimal power flow (OPF) analysis. This paper introduces a data-driven distributionally robust chance-constrained (DRCC) approach [...] Read more.
The increasing integration of distributed energy resources into distribution feeders introduces significant uncertainties, stemming from volatile renewable sources and other fluctuating electrical elements, which pose substantial challenges for optimal power flow (OPF) analysis. This paper introduces a data-driven distributionally robust chance-constrained (DRCC) approach to address the stochastic Alternating Current (AC) OPF problem in distribution grids, where the exact probability distributions of uncertainties are unknown. The proposed method utilizes the Wasserstein metric to construct an ambiguity set based on empirical distributions derived from historical data, eliminating the need for prior knowledge of the underlying probability distributions. Notably, the size of the Wasserstein ball within the ambiguity set is inversely related to the volume of available data, allowing for adaptive robustness. Moreover, a computationally efficient reformulation of the DRCC-OPF model is developed using the LinDistFlow AC power flow approximation. The effectiveness and precision of the developed method are validated through multiple IEEE distribution test cases, demonstrating higher reliability of the security constraints compared with other methods. As more data become available, this reliability is systematically and securely adjusted to achieve greater economic efficiency. Full article
Show Figures

Figure 1

Figure 1
<p>Illustrative example of the main notations in radial distribution grids.</p>
Full article ">Figure 2
<p>The flowchart of the sequence of operations.</p>
Full article ">Figure 3
<p>Topology diagram of IEEE 15-bus system.</p>
Full article ">Figure 4
<p>Total cost evolves with samples under different models.</p>
Full article ">Figure 5
<p>The operating costs for different parameters in <span class="html-italic">N</span> = 100, the radius of case 1 is (<a href="#FD16-electronics-14-00822" class="html-disp-formula">16</a>), and the radius of case 2 is (<a href="#FD17-electronics-14-00822" class="html-disp-formula">17</a>).</p>
Full article ">Figure 6
<p>Reliable of the IEEE 141-bus test system.</p>
Full article ">
20 pages, 516 KiB  
Article
Design of a Serendipity-Incorporated Recommender System
by Yuri Kim, Seoyeon Oh, Chaerin Noh, Eunbeen Hong and Seongbin Park
Electronics 2025, 14(4), 821; https://doi.org/10.3390/electronics14040821 - 19 Feb 2025
Abstract
Unexpected yet advantageous findings, often referred to as serendipitous discoveries, are becoming increasingly significant in the field of computer science. This research aims to examine the impact of factors that could potentially trigger such serendipity within a recommender system (RS) and consequently proposes [...] Read more.
Unexpected yet advantageous findings, often referred to as serendipitous discoveries, are becoming increasingly significant in the field of computer science. This research aims to examine the impact of factors that could potentially trigger such serendipity within a recommender system (RS) and consequently proposes a novel, serendipity-incorporated recommender system (SRS). The SRS is developed by integrating elements that could stimulate the occurrence of serendipity into an RS algorithm. These elements include interestingness, diversity, and unexpectedness. As a result, the SRS is equipped to provide users with recommendations that are surprising, intriguing, and atypical. The algorithm within the SRS recommends three items predicated on a user’s preferred item. To facilitate the selection of items to be recommended, we have designed a computation method called the ’serendipity measure’, which is tasked with calculating the weights of all items. Our innovative algorithm and its efficient execution are expounded upon extensively in this study. The performance of the SRS was assessed using a quantitative serendipity evaluation model (QSEM). This model is a quantitative tool designed to measure the probability of users encountering serendipitous events within a specific information space. We conducted a user study to compare the SRS with the traditional cold-start recommender system (CRS), and the feedback for the SRS was positively received. The experiments confirm the viability of cultivating a serendipitous environment from a system’s perspective. The test results also underline the exciting potential that serendipity brings to recommender systems. Full article
Show Figures

Figure 1

Figure 1
<p>Sample instances for recommender system algorithm.</p>
Full article ">Figure 2
<p>A simple demonstration of the similarity measure used in CRS.</p>
Full article ">Figure 3
<p>Three phases of the QSEM.</p>
Full article ">Figure 4
<p>Three phases of the E-QSEM.</p>
Full article ">Figure 5
<p>An illustration of unexpectedness calculation process with a simple example.</p>
Full article ">Figure 6
<p>Sample dataset of movie ratings (adapted from the MovieLens dataset [<a href="#B21-electronics-14-00821" class="html-bibr">21</a>]).</p>
Full article ">Figure 7
<p>Comparison of Average Ratings for CRS and SRS in User Survey Responses.</p>
Full article ">
34 pages, 442 KiB  
Review
A Review of Multi-Agent Reinforcement Learning Algorithms
by Jiaxin Liang, Haotian Miao, Kai Li, Jianheng Tan, Xi Wang, Rui Luo and Yueqiu Jiang
Electronics 2025, 14(4), 820; https://doi.org/10.3390/electronics14040820 - 19 Feb 2025
Abstract
In recent years, multi-agent reinforcement learning algorithms have demonstrated immense potential in various fields, such as robotic collaboration and game AI. This paper introduces the modeling concepts of single-agent and multi-agent systems: the fundamental principles of Markov Decision Processes and Markov Games. The [...] Read more.
In recent years, multi-agent reinforcement learning algorithms have demonstrated immense potential in various fields, such as robotic collaboration and game AI. This paper introduces the modeling concepts of single-agent and multi-agent systems: the fundamental principles of Markov Decision Processes and Markov Games. The reinforcement learning algorithms are divided into three categories: value-based, strategy-based, and actor–critic algorithms, and the algorithms and applications are introduced. Based on differences in reward functions, multi-agent reinforcement learning algorithms are further classified into three categories: fully cooperative, fully competitive, and mixed types. The paper systematically reviews and analyzes their basic principles, applications in multi-agent systems, challenges faced, and corresponding solutions. Specifically, it discusses the challenges faced by multi-agent reinforcement learning algorithms from four aspects: dimensionality, non-stationarity, partial observability, and scalability. Additionally, it surveys existing algorithm-training environments in the field of multi-agent systems and summarizes the applications of multi-agent reinforcement learning algorithms across different domains. Through this discussion, readers can gain a comprehensive understanding of the current research status and future trends in multi-agent reinforcement learning algorithms, providing valuable insights for further exploration and application in this field. Full article
(This article belongs to the Topic Agents and Multi-Agent Systems)
Show Figures

Figure 1

Figure 1
<p>Development stages of MARL.</p>
Full article ">Figure 2
<p>Basic framework of reinforcement learning.</p>
Full article ">Figure 3
<p>Classification of reinforcement learning algorithms.</p>
Full article ">Figure 4
<p>Summary of experimental platform.</p>
Full article ">
18 pages, 514 KiB  
Article
Incremental Repair Feedback on Automated Assessment of Programming Assignments
by José Carlos Paiva, José Paulo Leal and Álvaro Figueira
Electronics 2025, 14(4), 819; https://doi.org/10.3390/electronics14040819 - 19 Feb 2025
Abstract
Automated assessment tools for programming assignments have become increasingly popular in computing education. These tools offer a cost-effective and highly available way to provide timely and consistent feedback to students. However, when evaluating a logically incorrect source code, there are some reasonable concerns [...] Read more.
Automated assessment tools for programming assignments have become increasingly popular in computing education. These tools offer a cost-effective and highly available way to provide timely and consistent feedback to students. However, when evaluating a logically incorrect source code, there are some reasonable concerns about the formative gap in the feedback generated by such tools compared to that of human teaching assistants. A teaching assistant either pinpoints logical errors, describes how the program fails to perform the proposed task, or suggests possible ways to fix mistakes without revealing the correct code. On the other hand, automated assessment tools typically return a measure of the program’s correctness, possibly backed by failing test cases and, only in a few cases, fixes to the program. In this paper, we introduce a tool, AsanasAssist, to generate formative feedback messages to students to repair functionality mistakes in the submitted source code based on the most similar algorithmic strategy solution. These suggestions are delivered with incremental levels of detail according to the student’s needs, from identifying the block containing the error to displaying the correct source code. Furthermore, we evaluate how well the automatically generated messages provided by AsanasAssist match those provided by a human teaching assistant. The results demonstrate that the tool achieves feedback comparable to that of a human grader while being able to provide it just in time. Full article
(This article belongs to the Special Issue Program Slicing and Source Code Analysis: Methods and Applications)
30 pages, 6912 KiB  
Review
A Framework for Embedded Non-Volatile Memory Development: Examples from Pb(ZrxTi1−x)O3 Ferroelectric Memory Development at Texas Instruments
by Ted Moise, Scott Summerfelt and John Rodriguez
Electronics 2025, 14(4), 818; https://doi.org/10.3390/electronics14040818 - 19 Feb 2025
Abstract
An overview of the steps employed to advance non-volatile Pb(ZrxTi1−x)O3-based materials from parallel capacitor array test structures to embedded 130 nm (1.5 V operation) memory product release is presented. Specific development stages include parallel capacitor array evaluation, [...] Read more.
An overview of the steps employed to advance non-volatile Pb(ZrxTi1−x)O3-based materials from parallel capacitor array test structures to embedded 130 nm (1.5 V operation) memory product release is presented. Specific development stages include parallel capacitor array evaluation, capacitor characterization array development, memory macro creation and measurement, and initial product design and qualification. Representative data, learning goals, and critical outputs will be presented for each development phase. We note that the cost and complexity of the development effort increase dramatically as the new technology approaches high-volume manufacturing. We hope that the documentation of our experiences in this manuscript may be of assistance to those teams striving to create the next generations of non-volatile embedded memory technology. Full article
(This article belongs to the Special Issue Ferroelectric Materials and Applications)
Show Figures

Figure 1

Figure 1
<p>Comparison of write and read energy/bit for similar 3.3 V input/output, 4 Mb parallel access non-volatile products.</p>
Full article ">Figure 2
<p>An overview of the development stages used to advance from standalone capacitors to an initial product.</p>
Full article ">Figure 3
<p>Parallel capacitor arrays used to evaluate scaled PZT capacitor electrical and reliability properties using either three masks (<b>a</b>) or five masks (<b>b</b>). Adapted with permission from [<a href="#B77-electronics-14-00818" class="html-bibr">77</a>].</p>
Full article ">Figure 4
<p>Hysteresis loops for varying-thickness MOCVD PZT (<b>a</b>). Pulsed switched polarization for varying capacitor areas (<b>b</b>). Adapted with permission from [<a href="#B64-electronics-14-00818" class="html-bibr">64</a>,<a href="#B77-electronics-14-00818" class="html-bibr">77</a>].</p>
Full article ">Figure 5
<p>FRAM module integration cross-section. The FRAM module is a two-mask adder to the standard 130 nm CMOS process. Reprinted with permission from [<a href="#B145-electronics-14-00818" class="html-bibr">145</a>].</p>
Full article ">Figure 6
<p>4 Mb capacitor characterization array photograph and device features. Reprinted with permission from [<a href="#B64-electronics-14-00818" class="html-bibr">64</a>].</p>
Full article ">Figure 7
<p>Measured data 0 and data 1 polarizations for 144 kbit array at 1.5 V with a 100 ns cycle time. Reprinted with permission from [<a href="#B143-electronics-14-00818" class="html-bibr">143</a>].</p>
Full article ">Figure 8
<p>Endurance properties up to 1 × 10<sup>12</sup> cycles showing data 1 signal level increase relative to initial values. Adapted with permission from [<a href="#B146-electronics-14-00818" class="html-bibr">146</a>].</p>
Full article ">Figure 9
<p>FRAM data retention test procedure. A high-temperature “same-state (SS)” bake of a bit cell in one logic state can strengthen that state, while it weakens the capacitor’s ability to store data in the complement logic state (“opposite state (OS)”) because of imprint. The thermal depolarization bake step mimics the highest operating temperature of the product (typically 85 °C). Reprinted with permission from [<a href="#B150-electronics-14-00818" class="html-bibr">150</a>].</p>
Full article ">Figure 10
<p>Opposite-state data retention properties measured at 125 °C for 20 h and 1000 h bake durations. Adapted with permission from [<a href="#B143-electronics-14-00818" class="html-bibr">143</a>].</p>
Full article ">Figure 11
<p>An 8 Mbit FRAM optical micrograph and device features. Reprinted with permission from [<a href="#B145-electronics-14-00818" class="html-bibr">145</a>].</p>
Full article ">Figure 12
<p>Bit distributions for 8 Mb 1T-1C memory macro. Adapted with permission from [<a href="#B150-electronics-14-00818" class="html-bibr">150</a>].</p>
Full article ">Figure 13
<p>Signal distributions are shown comparing non-cycled and cycled bits after 5.4 × 10<sup>13</sup> cycles at 25 °C ambient temperature. No degradation in the intrinsic data margin window is observed after 5.4 × 10<sup>13</sup> cycles. Reprinted with permission from [<a href="#B155-electronics-14-00818" class="html-bibr">155</a>].</p>
Full article ">Figure 14
<p>Cycling wear-out rate is strongly accelerated by voltage. Cycling life is estimated to be greater than 5 × 10<sup>14</sup> cycles at 1.5 V. Reprinted with permission from [<a href="#B109-electronics-14-00818" class="html-bibr">109</a>].</p>
Full article ">Figure 15
<p>Graph of 8 Mb 1T-1C bit distributions following 1000 h bake at 125 °C. Reprinted with permission from [<a href="#B155-electronics-14-00818" class="html-bibr">155</a>].</p>
Full article ">Figure 16
<p>The lowest ‘1’ bit cell signal level is plotted as a function of bake time with bake temperature as a parameter. Based on these data, a 1.4 eV activation energy has been extracted. Reprinted with permission from [<a href="#B109-electronics-14-00818" class="html-bibr">109</a>].</p>
Full article ">Figure 17
<p>Data margin screens for reliability were established using extensive same-state and opposite-state data retention tests. The initial data margin is critical to ensure long-term memory reliability. Reprinted with permission from [<a href="#B109-electronics-14-00818" class="html-bibr">109</a>].</p>
Full article ">Figure 18
<p>The enthusiasm roller coaster describes the decade-long development journey from the initial demonstration of low-voltage PZT capacitors to the release of the first product.</p>
Full article ">
24 pages, 4908 KiB  
Article
Tensor-Based Uniform and Discrete Multi-View Projection Clustering
by Linlin Ma, Haomin Li, Wenke Zang, Xincheng Liu and Minghe Sun
Electronics 2025, 14(4), 817; https://doi.org/10.3390/electronics14040817 - 19 Feb 2025
Abstract
Multi-view graph clustering (MVGC) utilizes affinity graphs to efficiently obtain information between views. Although various excellent MVGC methods have been proposed, they still have many limitations. To surmount these limitations, this work develops a novel tensor-based unified and discrete multi-view projection clustering (TUDMPC) [...] Read more.
Multi-view graph clustering (MVGC) utilizes affinity graphs to efficiently obtain information between views. Although various excellent MVGC methods have been proposed, they still have many limitations. To surmount these limitations, this work develops a novel tensor-based unified and discrete multi-view projection clustering (TUDMPC) approach. Specifically, TUDMPC uses projection and the L2,1-norm for feature selection to reduce the effects of redundancy and noise. Meanwhile, the differences among similar graphs are minimized through the tensor kernel norm to better leverage information across views and capture high-order correlations. In addition, the rank constraint is applied to keep the affinity graphs with a discrete cluster structure, and the clustering results are obtained directly in a unified joint framework. Finally, an efficient optimization algorithm is proposed to obtain the clustering results. Experiments are conducted to compare the clustering results of TUDMPC with seven baseline methods. The results show that TUDMPC outperforms the existing methods. Full article
(This article belongs to the Special Issue Emerging Distributed/Parallel Computing Systems)
Show Figures

Figure 1

Figure 1
<p>Flowchart of the TUDMPC algorithm.</p>
Full article ">Figure 2
<p>Visualization of the affinity matrices of the MSRC_v1 dataset. (<b>a</b>) SC. (<b>b</b>) MCGC. (<b>c</b>) MVGL. (<b>d</b>) GMC. (<b>e</b>) SFMC. (<b>f</b>) TUDMPC.</p>
Full article ">Figure 3
<p>Visualization of the affinity matrices of the NGs dataset. (<b>a</b>) SC. (<b>b</b>) MCGC. (<b>c</b>) MVGL. (<b>d</b>) GMC. (<b>e</b>) SFMC. (<b>f</b>) TUDMPC.</p>
Full article ">Figure 4
<p>Visualization of the affinity matrices of the 100leaves dataset. (<b>a</b>) SC. (<b>b</b>) MCGC. (<b>c</b>) MVGL. (<b>d</b>) GMC. (<b>e</b>) SFMC. (<b>f</b>) TUDMPC.</p>
Full article ">Figure 5
<p>Visualization of the HW2sources dataset. (<b>a</b>) SC. (<b>b</b>) Co-regMSC. (<b>c</b>) AWP. (<b>d</b>) MCGC. (<b>e</b>) MVGL. (<b>f</b>) TUDMPC.</p>
Full article ">Figure 6
<p>Visualization of the MSRC_v1 dataset. (<b>a</b>) SC. (<b>b</b>) Co-regMSC. (<b>c</b>) AWP. (<b>d</b>) MCGC. (<b>e</b>) MVGL. (<b>f</b>) TUDMPC.</p>
Full article ">Figure 7
<p>Some face images from the ORL dataset (10 × 10).</p>
Full article ">Figure 8
<p>Some handwritten digital images from the HW dataset (10 × 50).</p>
Full article ">Figure 9
<p>Some face image recognition results of TUDMPC on the ORL dataset (10 × 10).</p>
Full article ">Figure 10
<p>Some handwritten digital image recognition results of TUDMPC on the HW dataset (10 × 50).</p>
Full article ">Figure 11
<p>Results of the ablation experiments on the 100leaves dataset.</p>
Full article ">Figure 12
<p>Convergence on some datasets. (<b>a</b>) MRSC_v1. (<b>b</b>) HW2source. (<b>c</b>) ORL. (<b>d</b>) 100leaves.</p>
Full article ">Figure 13
<p>Sensitivity analysis on different datasets as parameters <math display="inline"><semantics> <mi>α</mi> </semantics></math> and <span class="html-italic">k</span> change. (<b>a</b>) MSRC_v1. (<b>b</b>) HW2sources. (<b>c</b>) 100leaves. (<b>d</b>) NGs. (<b>e</b>) ORL.</p>
Full article ">Figure 14
<p>Results of the Nemenyi Test. (<b>a</b>) ACC. (<b>b</b>) NMI. (<b>c</b>) Purity.</p>
Full article ">
28 pages, 4266 KiB  
Article
Hierarchical Vision–Language Pre-Training with Freezing Strategy for Multi-Level Semantic Alignment
by Huiming Xie, Yang Qin and Shuxue Ding
Electronics 2025, 14(4), 816; https://doi.org/10.3390/electronics14040816 - 19 Feb 2025
Abstract
Vision–language pre-training (VLP) faces challenges in aligning hierarchical textual semantics (words/phrases/sentences) with multi-scale visual features (objects/relations/global context). We propose a hierarchical VLP model (HieVLP) that addresses such challenges through semantic decomposition and progressive alignment. Textually, a semantic parser deconstructs captions into word-, phrase-, [...] Read more.
Vision–language pre-training (VLP) faces challenges in aligning hierarchical textual semantics (words/phrases/sentences) with multi-scale visual features (objects/relations/global context). We propose a hierarchical VLP model (HieVLP) that addresses such challenges through semantic decomposition and progressive alignment. Textually, a semantic parser deconstructs captions into word-, phrase-, and sentence-level components, which are encoded via hierarchical BERT layers. Visually, a Swin Transformer extracts object- (local), relation- (mid-scale), and global-level features through shifted window hierarchies. During pre-training, a freezing strategy sequentially activates text layers (sentence→phrase→word), aligning each with the corresponding visual scales via contrastive and language modeling losses. The experimental evaluations demonstrate that HieVLP outperforms hierarchical baselines across various tasks, with the performance improvements ranging from approximately 3.2% to 11.2%. In the image captioning task, HieVLP exhibits an average CIDEr improvement of around 7.2% and a 2.1% improvement in the SPICE metric. For image–text retrieval, it achieves recall increases of 4.7–6.8%. In reasoning tasks, HieVLP boosts accuracy by 2.96–5.8%. These results validate that explicit multi-level alignment enables contextually coherent caption generation and precise cross-modal reasoning. Full article
(This article belongs to the Section Artificial Intelligence)
Show Figures

Figure 1

Figure 1
<p>Multi-level semantic alignment visualization: Object–word correspondences (purple), relation–phrase interactions (green), and global–sentence contexts (red). Subfigures (<b>a</b>,<b>b</b>) are detected under a standard textual description, with their individual targets exhibiting notable scale variations.</p>
Full article ">Figure 2
<p>Overall architecture of the proposed model. The HieVLP model employs a hierarchical structure alongside a fixed strategy to bolster the alignment process. The proposed strategy facilitates a multi-level amalgamation of semantic information from both visual and textual domains, thereby enhancing the pre-training performance.</p>
Full article ">Figure 3
<p>Comparison of Gradient Norm Evolution with and without Freezing Strategy.</p>
Full article ">Figure 4
<p>Comparison of Validation Loss with and without Freezing Strategy.</p>
Full article ">Figure 5
<p>Performance Comparison between Freezing Strategy and Other Regularization Methods.</p>
Full article ">Figure 6
<p>Visualization of qualitative examples on the COCO dataset. The first column indicates images from the COCO validation set. The second column shows the five human-annotated ground-truth captions. The third column indicates captions generated by our pre-trained HieVLP and baseline models and the corresponding CIDEr scores (CIDEr scores are in parentheses in the third column). The baseline model is the one with the highest CIDEr score for generating descriptions under the current image, selected from all the compared models in <a href="#sec4dot4dot3-electronics-14-00816" class="html-sec">Section 4.4.3</a>. We extract the hierarchical information of the generated descriptions using the word parser, phrase parser, and sentence parser, respectively. Intuitively, our model descriptions compare the ground-truth and baseline models that contain more multi-layered information, higher CIDEr scores, more long-term structure, and more accurate and detailed descriptions.</p>
Full article ">
19 pages, 7903 KiB  
Article
Fast Temperature Calculation Method for Spindle Servo Permanent Magnet Motors Under Full Operating Conditions Based on the Thermal Network Method
by Sheng Ma, Yijia Li, Xueyan Hao, Bo Zhang and Wei Feng
Electronics 2025, 14(4), 815; https://doi.org/10.3390/electronics14040815 - 19 Feb 2025
Abstract
In CNC machines, the temperature field analysis of spindle servo permanent magnet motors (SSPMMs) under rated load, overload, and weak magnetic conditions is critical for ensuring stable operation and machining accuracy. This paper proposes a temperature calculation method for SSPMMs based on the [...] Read more.
In CNC machines, the temperature field analysis of spindle servo permanent magnet motors (SSPMMs) under rated load, overload, and weak magnetic conditions is critical for ensuring stable operation and machining accuracy. This paper proposes a temperature calculation method for SSPMMs based on the thermal network method, which is used to quickly evaluate the temperature performance of SSPMMs under different operating conditions during design. This method can calculate the steady-state or transient temperature rise under different operating conditions. First, the electromagnetic performance and heat sources of the SSPMMs were analyzed. Then, based on the thermal network method, the equivalent thermal resistances and equivalent heat dissipation coefficients of the motor components were calculated. By iterating the heat balance equation or solving the heat conduction equation for different operating conditions, the temperature distribution of SSPMMs under different operating conditions was obtained. The accuracy of the thermal network model was validated through temperature analysis using fluid–structure interaction simulations and prototype testing. The results show that the relative error between the winding temperature calculated by the proposed equivalent thermal network model and the measured temperature under different operating conditions is less than 5%. This paper provides a theoretical basis for the thermal management of SSPMM, which can quickly and accurately evaluate the temperature rise in the motor during design. Full article
Show Figures

Figure 1

Figure 1
<p>Structure of water-cooled 12-pole 54-slot SSPMM.</p>
Full article ">Figure 2
<p>Efficiency map of SSPMM.</p>
Full article ">Figure 3
<p>The thermal network structure of SSPMM.</p>
Full article ">Figure 4
<p>The sub-thermal network structure of SSPMM. (<b>a</b>) node 30; (<b>b</b>) node 8; (<b>c</b>) node 17.</p>
Full article ">Figure 5
<p>SSPMM temperature calculation flowchart based on thermal network method.</p>
Full article ">Figure 6
<p>Three-dimensional model of SSPMM. (<b>a</b>) Spiral waterway structure; (<b>b</b>) three-dimensional structure section view.</p>
Full article ">Figure 7
<p>Temperature distributions in SSPMM under rated condition: (<b>a</b>) stator; (<b>b</b>) stator winding; (<b>c</b>) permanent magnet; and (<b>d</b>) cooling water.</p>
Full article ">Figure 8
<p>Temperature distributions in SSPMM under high speed condition: (<b>a</b>) stator; (<b>b</b>) stator winding; (<b>c</b>) permanent magnet; and (<b>d</b>) cooling water.</p>
Full article ">Figure 9
<p>Prototype temperature test: (<b>a</b>) motor test system; (<b>b</b>) mechanical back-to-back test bench.</p>
Full article ">Figure 10
<p>Test temperature curves under rated conditions.</p>
Full article ">Figure 11
<p>Test temperature curves under overload condition.</p>
Full article ">
25 pages, 5090 KiB  
Article
Research on Intelligent Verification of Equipment Information in Engineering Drawings Based on Deep Learning
by Zicheng Zhang and Yurou He
Electronics 2025, 14(4), 814; https://doi.org/10.3390/electronics14040814 - 19 Feb 2025
Abstract
This paper focuses on the crucial task of automatic recognition and understanding of table structures in engineering drawings and document processing. Given the importance of tables in information display and the urgent need for automated processing of tables in the digitalization process, an [...] Read more.
This paper focuses on the crucial task of automatic recognition and understanding of table structures in engineering drawings and document processing. Given the importance of tables in information display and the urgent need for automated processing of tables in the digitalization process, an intelligent verification method is proposed. This method integrates multiple key techniques: YOLOv10 is used for table object recognition, achieving a precision of 0.891, a recall rate of 0.899, mAP50 of 0.922, and mAP50-95 of 0.677 in table recognition, demonstrating strong target detection capabilities; the improved LORE algorithm is adopted to extract table structures, breaking through the limitations of the original algorithm by segmenting large-sized images, with a table extraction accuracy rate reaching 91.61% and significantly improving the accuracy of handling complex tables; RapidOCR is utilized to achieve text recognition and cell correspondence, solving the problem of text-cell matching; for equipment name semantic matching, a method based on BERT is introduced and calculated using a comprehensive scoring method. Meanwhile, an improved cuckoo search algorithm is proposed to optimize the adjustment factors, avoiding local optima through sine optimization and the catfish effect. Experiments show the accuracy of equipment name matching in semantic similarity calculation approaches 100%. Finally, the paper provides a concrete system practice to prove the effectiveness of the algorithm. In conclusion, through experimental comparisons, this method exhibits excellent performance in table area location, structure recognition, and semantic matching and is of great significance and practical value in advancing table data processing technology in engineering drawings. Full article
(This article belongs to the Section Artificial Intelligence)
Show Figures

Figure 1

Figure 1
<p>The framework of intelligent verification methods.</p>
Full article ">Figure 2
<p>The framework of YOLOv10.</p>
Full article ">Figure 3
<p>Illustration of the improved LORE algorithm.</p>
Full article ">Figure 4
<p>First-last layer average pooling.</p>
Full article ">Figure 5
<p>Improved cuckoo search algorithm flowchart.</p>
Full article ">Figure 6
<p>Iteration curve of algorithm training effectiveness.</p>
Full article ">Figure 7
<p>Display of recognition results.</p>
Full article ">Figure 8
<p>Comparison of the recognition process of this paper’s algorithm with the original LORE algorithm.</p>
Full article ">Figure 9
<p>Iteration curves of the three functions for CS and ICS. (<b>a</b>) Iteration Curves of Function F1 (<b>b</b>) Iteration Curves of Function F2 (<b>c</b>) Iteration Curves of Function F3.</p>
Full article ">Figure 10
<p>Iteration curves of the three algorithms.</p>
Full article ">Figure 11
<p>Schematic diagram of system process, model and components.</p>
Full article ">Figure 12
<p>Matching result system screenshot.</p>
Full article ">
8 pages, 17999 KiB  
Article
4 × 4 Wideband Slot Antenna Array Fed by TE440 Mode Based on Groove Gap Waveguide
by Yuanjun Shen, Tianling Zhang, Liangqin Luo, Honghuan Zhu and Lei Chen
Electronics 2025, 14(4), 813; https://doi.org/10.3390/electronics14040813 - 19 Feb 2025
Abstract
A 4 × 4 wideband millimeter-wave (mmWave) slot array antenna excited by the TE440 mode based on the groove gap waveguide is presented in this paper. A vertical waveguide located in the center of the cavity and two ridges are used to [...] Read more.
A 4 × 4 wideband millimeter-wave (mmWave) slot array antenna excited by the TE440 mode based on the groove gap waveguide is presented in this paper. A vertical waveguide located in the center of the cavity and two ridges are used to excite the TE440 mode. In addition, a pair of corrugations acting as the soft surface are added on the top of the array antenna to improve the gain. A 4 × 4 prototype is fabricated and measured. The measured and simulated results are in great agreement. The measured results show that the proposed array antenna achieved an impedance bandwidth (|S11| < −10 dB) of 26.7% from 26.14 to 34.2 GHz, and the maximum gain is 17.7 dBi. The proposed array antenna avoids the complicated feeding network, allowing us to reduce the manufacturing cost. Full article
(This article belongs to the Special Issue Antenna and Array Design for Future Sensing and Communication System)
Show Figures

Figure 1

Figure 1
<p>The geometry of the 4 × 4 slot array antenna introduced using TE<sub>440</sub>.</p>
Full article ">Figure 2
<p>The dispersion diagram of the unit cell.</p>
Full article ">Figure 3
<p>The geometry of the metal plates (in mm): (<b>a</b>) top plate and (<b>b</b>) bottom plate.</p>
Full article ">Figure 4
<p>The simulated E-field distribution within the GGW cavity without slots at 31.62 GHz.</p>
Full article ">Figure 5
<p>The evolution process of the newly introduced wideband 4 × 4 slot array antenna.</p>
Full article ">Figure 6
<p>The simulated results of the proposed 4 × 4 wideband slot array antenna after Steps 1–3: (<b>a</b>) reflection coefficient and (<b>b</b>) gain.</p>
Full article ">Figure 7
<p>The simulated maximum E-field distribution over the antenna’s top surface at 29 GHz: (<b>a</b>) without ridges and corrugations, (<b>b</b>) with ridges but without corrugations, and (<b>c</b>) with both ridges and corrugations.</p>
Full article ">Figure 8
<p>Photographs of the prototype: (<b>a</b>) from the oblique top, (<b>b</b>) from the oblique top (inside), and (<b>c</b>) from the bottom.</p>
Full article ">Figure 9
<p>The measured and simulated reflection coefficients of the newly introduced 4 × 4 wideband slot array antenna.</p>
Full article ">Figure 10
<p>The measured and simulated radiation patterns of the proposed 4 × 4 wideband slot array antenna: (<b>a</b>) 27 GHz, (<b>b</b>) 30.5 GHz, and (<b>c</b>) 34 GHz.</p>
Full article ">Figure 11
<p>The measured and simulated gains of the newly introduced slot array antenna.</p>
Full article ">
17 pages, 9626 KiB  
Article
Semantic Segmentation of Distribution Network Point Clouds Based on NF-PTV2
by Long Han, Bin Song, Shaocheng Wu, Deyu Nie, Zhenyang Chen and Linong Wang
Electronics 2025, 14(4), 812; https://doi.org/10.3390/electronics14040812 - 19 Feb 2025
Abstract
An on-site survey is the primary task of working live in distribution networks. However, the traditional manual on-site survey method is not only not very intuitive but also inefficient. The application of 3D point cloud technology has opened up new avenues for on-site [...] Read more.
An on-site survey is the primary task of working live in distribution networks. However, the traditional manual on-site survey method is not only not very intuitive but also inefficient. The application of 3D point cloud technology has opened up new avenues for on-site surveys in life working in distribution networks. This paper focused on the application of the Point Transformer V2(PTV2) model in the segmentation of distribution network point clouds. Given its deficiencies in boundary discrimination ability and limited feature extraction ability when processing the point clouds of distribution networks, an improved Non-local Focal Loss-Point Transformer V2 (NF-PTV2) model was proposed. With PTV2 as its core, this model incorporated the Non-Local attention to capturing long-distance feature dependencies, thereby compensating for the PTV2 model’s shortcomings in extracting features of large-scale objects with complex features. Simultaneously, the Focal Loss function was introduced to address the issue of class imbalance and enhance the model’s learning ability for small complex samples. The experimental results demonstrated that the overall accuracy (OA) of this model on the distribution network dataset reached 93.28%, the mean intersection over union (mIoU) reached 81.58%, and the mean accuracy (mAcc) reached 87.21%. In summary, the NF-PTV2 model proposed in this article demonstrated good performance in the point cloud segmentation task of the distribution network and can accurately identify various objects, which, to some extent, overcomes the limitations of the PTV2 model. Full article
Show Figures

Figure 1

Figure 1
<p>Non-Local attention.</p>
Full article ">Figure 2
<p>The framework of the NF-PTV2 model. The red parts represent the modules that are unique to the NF-PTV2 model and are not included in the PTV2 model.</p>
Full article ">Figure 3
<p>Segmentation results of the NF-PTV2 model on the distribution network point cloud dataset. (<b>a</b>) shows the point cloud segmented by the NF-PTV2 model; (<b>b</b>) shows the manually labeled point cloud. The circles mark the areas where the NF-PTV2 model’s segmentation results differ from the manually labeled point clouds.</p>
Full article ">Figure 4
<p>Segmentation results of models embedded with an attention mechanism on the distribution network point cloud dataset.</p>
Full article ">Figure 5
<p>Classification of various models in Area 1 to Area 4. The circles in the figure mark the areas where differences exist between the results of various models.</p>
Full article ">Figure 6
<p>Test results of various models in Area 3 and Area 4 of the test set. The circles in the figure mark the areas where differences exist between the results of various models.</p>
Full article ">Figure 7
<p>The overall structure of the proposed method.</p>
Full article ">
25 pages, 20387 KiB  
Article
Connecting Visual Data to Privacy: Predicting and Measuring Privacy Risks in Images
by Hongpu Jiang, Jinxin Zuo and Yueming Lu
Electronics 2025, 14(4), 811; https://doi.org/10.3390/electronics14040811 - 19 Feb 2025
Abstract
More and more users openly share their information on online websites, with the resulting privacy issues being under scrutiny. Content such as a user’s personal data and location information is often asked for before posting to enforce the user’s privacy preferences; however, little [...] Read more.
More and more users openly share their information on online websites, with the resulting privacy issues being under scrutiny. Content such as a user’s personal data and location information is often asked for before posting to enforce the user’s privacy preferences; however, little attention has been paid to the lack of content (e.g., images) posted by the user. Even if privacy preferences are requested before images are published, publishers often remain unaware of the extent of privacy leakage associated with their data. To this end, we provide an image privacy metric scheme that incorporates users’ privacy preferences, with the core idea of assisting users in making data publishing decisions. First, we propose privacy-specific spatial attention mechanisms that can effectively improve the prediction accuracy. Next, we integrate set pair analysis (SPA) theory and use the network output as the privacy value. Finally, we combine a user study to understand the privacy preferences of different users with respect to these attributes and combine it with principal component analysis to correct and enforce user privacy preferences. Our model is trained with the ability to predict privacy risk end-to-end, thus being able to guide the user in sharing data in open platforms. We use the image privacy dataset, VISPR, to predict privacy information better than other methods. Full article
(This article belongs to the Topic Recent Advances in Security, Privacy, and Trust)
Show Figures

Figure 1

Figure 1
<p>Examples of user-generated content.</p>
Full article ">Figure 2
<p>Employing automated deep learning and set pair analysis techniques to enhance privacy measurement practices on social media platforms.</p>
Full article ">Figure 3
<p>The structure of privacy-specific spatial attention.</p>
Full article ">Figure 4
<p>Intuitive diagrams of identity/discrepancy/contrary relationships in privacy risk assessment based on set pair analysis.</p>
Full article ">Figure 5
<p>Average precision and recall scores.</p>
Full article ">Figure 6
<p>Set pair analysis.</p>
Full article ">
25 pages, 2916 KiB  
Article
Improving Cyber Defense Against Ransomware: A Generative Adversarial Networks-Based Adversarial Training Approach for Long Short-Term Memory Network Classifier
by Ping Wang, Hsiao-Chung Lin, Jia-Hong Chen, Wen-Hui Lin and Hao-Cyuan Li
Electronics 2025, 14(4), 810; https://doi.org/10.3390/electronics14040810 - 19 Feb 2025
Abstract
The rapid proliferation of ransomware variants necessitates more effective detection mechanisms, as traditional signature-based methods are increasingly inadequate. These conventional methods rely on manual feature extraction and matching, which are time-consuming and limited to known threats. This study addresses the escalating challenge of [...] Read more.
The rapid proliferation of ransomware variants necessitates more effective detection mechanisms, as traditional signature-based methods are increasingly inadequate. These conventional methods rely on manual feature extraction and matching, which are time-consuming and limited to known threats. This study addresses the escalating challenge of ransomware threats in cybersecurity by proposing a novel deep learning model, LSTM-EDadver, which leverages Generative Adversarial Networks (GANs) and Carlini and Wagner (CW) attacks to enhance malware detection capabilities. LSTM-EDadver innovatively generates adversarial examples (AEs) using sequential features derived from ransomware behaviors, thus training deep learning models to improve their robustness and accuracy. The methodology combines Cuckoo sandbox analysis with conceptual lattice ontology to capture a wide range of ransomware families and their variants. This approach not only addresses the shortcomings of existing models but also simulates real-world adversarial conditions during the validation phase by subjecting the models to CW attacks. The experimental results demonstrate that LSTM-EDadver achieves a classification accuracy of 96.59%. This performance was achieved using a dataset of 1328 ransomware samples (across 32 ransomware families) and 519 normal instances, outperforming traditional RNN, LSTM, and GCU models, which recorded accuracies of 90.01%, 93.95%, and 94.53%, respectively. The proposed model also shows significant improvements in F1-score, ranging from 2.49% to 6.64% compared to existing models without adversarial training. This advancement underscores the effectiveness of integrating GAN-generated attack command sequences into model training. Full article
(This article belongs to the Section Networks)
Show Figures

Figure 1

Figure 1
<p>The evolution of malware detection methods.</p>
Full article ">Figure 2
<p>Diagram of the developed malware classification model.</p>
Full article ">Figure 3
<p>The overall structure of the LSTM-ED<sub>adver</sub> model for ransomware detection.</p>
Full article ">Figure 4
<p>A deep learning model with auto-encoder structure for adversarial training.</p>
Full article ">Figure 5
<p>Subset Tree of Feature sets using the C4.5 algorithm.</p>
Full article ">Figure 6
<p>Concept lattice of 32 ransomware families.</p>
Full article ">Figure 7
<p>Classification accuracy.</p>
Full article ">Figure 8
<p>Loss value.</p>
Full article ">Figure 9
<p>Confusion matrix for LSTM-ED<sub>adver</sub>.</p>
Full article ">
23 pages, 1209 KiB  
Article
Towards a Multi-Objective and Contextual Multi-Criteria Recommender System for Enhancing User Well-Being in Sustainable Smart Homes
by Oumaima Stitini and Soulaimane Kaloun
Electronics 2025, 14(4), 809; https://doi.org/10.3390/electronics14040809 - 19 Feb 2025
Abstract
Smart homes have become an important part of our daily lives, changing our habits to make them easier to live with in a sustainable way. This study highlights a context-sensitive system that continuously adapts to the user’s current activities and physiological habits in [...] Read more.
Smart homes have become an important part of our daily lives, changing our habits to make them easier to live with in a sustainable way. This study highlights a context-sensitive system that continuously adapts to the user’s current activities and physiological habits in order to preserve physical and mental health while achieving sustainability goals. The system uses Internet of Things (IoT) sensors and smart home devices to measure indicators such as physical activity, heart rate, stress levels, and sleep quality. Based on these real-time measurements, the device offers personalized recommendations for a healthier lifestyle, such as physical activity reminders, stress management techniques, and sleep quality adjustments. By balancing the novelty and precision of its recommendations, the model aims to actively involve users without overloading them, thus promoting gradual and lasting behavioral changes. The architecture also incorporates multi-criteria evaluation measures, including accuracy and novelty-based diversity, to ensure an optimized user experience that is both accurate and adaptable. This paper proposes an advanced recommendation system for enhanced health monitoring by integrating multi-criteria decision-making and contextual awareness in order to have multi-objective results. The proposed system makes the personal recommendations with dynamic user categorization, using different kinds of notifications: reminder to exercise, monitoring heart health, and how to handle the stress. This approach is targeted to be scalable and adaptive to real-world conditions to enhance the overall effectiveness of health recommendations and strategies for preventive healthcare. The use of IoT will help in presenting a sound framework for personalized health interventions, enabling user engagement and better health outcomes. Full article
Show Figures

Figure 1

Figure 1
<p>Different filtering techniques used in recommender systems.</p>
Full article ">Figure 2
<p>Key aspects of context in CARS.</p>
Full article ">Figure 3
<p>Empowering smart homes: the role of context-aware systems and AI in personalized well-being.</p>
Full article ">Figure 4
<p>Context-aware algorithm types.</p>
Full article ">Figure 5
<p>Striking the balance: fusing novelty and accuracy in recommendations for sustainable well-being.</p>
Full article ">Figure 6
<p>Holistic flow for intelligent well-being: a smart ecosystem framework.</p>
Full article ">Figure 7
<p>Feature importances from different models.</p>
Full article ">Figure 8
<p>Comparative analysis of IoT device metrics: accuracy vs novelty.</p>
Full article ">
21 pages, 5976 KiB  
Article
Human-Centric Microgrid Optimization: A Two-Time-Scale Framework Integrating Consumer Behavior
by Ke Zeng, Hanqing Yang, Tieshan Li and Yue Long
Electronics 2025, 14(4), 808; https://doi.org/10.3390/electronics14040808 - 19 Feb 2025
Abstract
This paper presents a two-time-scale human-centric microgrid optimization framework, developed based on singular perturbation theory. First, a comprehensive model is constructed, integrating the electrical characteristics of microgrid components with the evolutionary dynamics of consumer behavior. Subsequently, the system is decomposed into distinct fast [...] Read more.
This paper presents a two-time-scale human-centric microgrid optimization framework, developed based on singular perturbation theory. First, a comprehensive model is constructed, integrating the electrical characteristics of microgrid components with the evolutionary dynamics of consumer behavior. Subsequently, the system is decomposed into distinct fast and slow time scales using singular perturbation theory, enabling the effective separation of rapid electrical responses from the slower motivational dynamics of consumer behavior. Tailored optimal control strategies are then formulated for each time scale to ensure the rapid stabilization of fast system dynamics in response to transient disturbances and the gradual optimization of slow system dynamics under steady-state conditions. Finally, the proposed approach is validated through preliminary numerical simulations, which demonstrate its potential effectiveness in maintaining microgrid stability under transient conditions, facilitating behavioral adaptation, and improving operational efficiency of the microgrid. Full article
(This article belongs to the Special Issue Innovations in Intelligent Microgrid Operation and Control)
Show Figures

Figure 1

Figure 1
<p>Two-time-scale microgrid framework: integrating consumer behavior with load and generation dynamics.</p>
Full article ">Figure 2
<p>Schematic representation of the internal electrical configuration of a prosumer in a microgrid.</p>
Full article ">Figure 3
<p>The optimization control process flowchart.</p>
Full article ">Figure 4
<p>The topology of interconnected prosumers.</p>
Full article ">Figure 5
<p>Scenario 1. Electrical Response of the System.</p>
Full article ">Figure 6
<p>Scenario 1. Dynamics of Consumer Behavior, Motivation, and Social Intervention.</p>
Full article ">Figure 7
<p>Scenario 2. Electrical Response of the System under Hedonic Value-Dominated Consumer.</p>
Full article ">Figure 8
<p>Scenario 2. Dynamics of Consumer Behavior, Motivation, and Social Intervention under Hedonic Value-Dominated Consumer.</p>
Full article ">Figure 9
<p>Scenario 2. Electrical Response of the System under Biospheric Value-Dominated Consumer.</p>
Full article ">Figure 10
<p>Scenario 2. Dynamics of Consumer Behavior, Motivation, and Social Intervention under Biospheric Value-Dominated Consumer.</p>
Full article ">Figure 11
<p>Scenario 3. Electrical Response of the System.</p>
Full article ">Figure 12
<p>Scenario 3. Dynamics of Consumer Behavior, Motivation, and Social Intervention.</p>
Full article ">
22 pages, 5289 KiB  
Article
Design of the New Dual-Polarized Broadband Phased Array Feed Antenna for the Sardinia Radio Telescope
by Paolo Maxia, Giovanni Andrea Casula, Alessandro Navarrini, Tonino Pisanu, Giuseppe Valente, Giacomo Muntoni and Giorgio Montisci
Electronics 2025, 14(4), 807; https://doi.org/10.3390/electronics14040807 - 19 Feb 2025
Abstract
High-sensitivity and large-scale surveys are essential in advancing radio astronomy, enabling detailed exploration of the universe. A Phased Array Feed (PAF) installed in the focal plane of a radio telescope significantly enhances mapping efficiency by increasing the instantaneous Field of View (FoV) and [...] Read more.
High-sensitivity and large-scale surveys are essential in advancing radio astronomy, enabling detailed exploration of the universe. A Phased Array Feed (PAF) installed in the focal plane of a radio telescope significantly enhances mapping efficiency by increasing the instantaneous Field of View (FoV) and improving sky sampling capabilities. This paper presents the design and optimization of a novel C-Band Phased Array Feed antenna for the Sardinia Radio Telescope (SRT). The system features an 8 × 8 array of dual-polarized elements optimized to achieve a uniform beam pattern and an edge taper of approximately 5 dB for single radiating elements within the 3.0–7.7 GHz frequency range. The proposed antenna addresses key efficiency limitations identified in the PHAROS 2 (PHased Arrays for Reflector Observing Systems) system, including the under-illumination of the Sardinia Radio Telescope’s primary mirror caused by narrow sub-array radiation patterns. By expanding the operational bandwidth and refining the radiation characteristics, this new design enables significantly improved performance across the broader frequency range of 3.0–7.7 GHz, enhancing the telescope’s capability for wide-field, high-resolution observations. Full article
(This article belongs to the Special Issue Microwave Devices: Analysis, Design, and Application)
Show Figures

Figure 1

Figure 1
<p>Simplified layout of the 64-element array, showing the central 32 active elements highlighted in orange.</p>
Full article ">Figure 2
<p>Diagram of a typical Linear Tapered Slot Antenna with key parameters, such as taper length (<span class="html-italic">L<sub>t</sub></span>), opening angle α, taper width (<span class="html-italic">W<sub>t</sub></span>), and feed slot width (<span class="html-italic">W<sub>s</sub></span>), clearly marked.</p>
Full article ">Figure 3
<p>Spacing between adjacent elements of the array.</p>
Full article ">Figure 4
<p>(<b>a</b>) Unit cell; (<b>b</b>) parameters of the feeding network.</p>
Full article ">Figure 5
<p>Implementation of the feeding probe integrated in the antenna.</p>
Full article ">Figure 6
<p>(<b>a</b>) 3D view of the 128-element array; (<b>b</b>) antenna feeding scheme for edge taper adjustment (in red color the active zone): the antenna numbered with 93 and highlighted in green is the only one that is powered.</p>
Full article ">Figure 7
<p>H-plane 5 dB edge taper: variation of taper length <span class="html-italic">L<sub>t</sub></span>.</p>
Full article ">Figure 8
<p>E-plane 5 dB edge taper: variation of taper length <span class="html-italic">L<sub>t</sub></span>.</p>
Full article ">Figure 9
<p>Single Linear Tapered Slot Antenna H-plane-normalized radiated field.</p>
Full article ">Figure 10
<p>Single Linear Tapered Slot Antenna E-plane-normalized radiated field.</p>
Full article ">Figure 11
<p>The −5 dB H-plane edge taper comparison.</p>
Full article ">Figure 12
<p>The −5 dB E-plane edge taper comparison.</p>
Full article ">Figure 13
<p>(<b>a</b>) Overall dimensions of the optimized unit cell, and (<b>b</b>) optimized parameters of the “L-Shaped” slot: all dimensions are in millimeters.</p>
Full article ">Figure 14
<p>S-parameters of the dual-polarized unit cell.</p>
Full article ">Figure 15
<p>(<b>a</b>) Finite array model with additional boundary elements, and (<b>b</b>) the antenna feeding scheme for the final configuration of the Phased Array antenna: the 16 passive additional elements are highlighted in blue color. In red color is the active zone.</p>
Full article ">Figure 16
<p>Comparison between the Active Reflection Coefficients of the single antenna of the unit cell and the antenna labelled with number 93 in the finite array.</p>
Full article ">Figure 17
<p>Simulated mutual coupling between some elements having the same polarization.</p>
Full article ">Figure 18
<p>Simulated mutual coupling between some elements having cross-polarization.</p>
Full article ">
16 pages, 393 KiB  
Article
Towards Effective Guidance of Smart Contract Fuzz Testing Based on Static Analysis
by Jeongwon Park and Jaeseung Choi
Electronics 2025, 14(4), 806; https://doi.org/10.3390/electronics14040806 - 19 Feb 2025
Abstract
In smart contract fuzz testing, it is crucial to consider the inter-dependencies between the contract functions. To effectively test the business logic of a contract, its functions must be invoked in a meaningful order. In this paper, we propose techniques that utilize static [...] Read more.
In smart contract fuzz testing, it is crucial to consider the inter-dependencies between the contract functions. To effectively test the business logic of a contract, its functions must be invoked in a meaningful order. In this paper, we propose techniques that utilize static analysis on Ethereum bytecode to tackle this challenge. When compared with the current state-of-the-art, our approach takes Solidity compiler’s variable packing optimization into account and allows more precise analysis of the data-flows between functions. In addition, we devise a novel test case initialization algorithm for fuzz testing, which minimizes the redundancy in the generated seed set. Our algorithm reduces test cases that share similar function call patterns and leads to more effective testing of the contract code during the fuzz testing. Experimental results show that the proposed techniques improve the effectiveness of smart contract fuzz testing for vulnerability detection. Specifically, our techniques enabled the fuzz testing tool to trigger the target bugs in the benchmark 3.0 times faster on average. Full article
(This article belongs to the Section Computer Science & Engineering)
Show Figures

Figure 1

Figure 1
<p>Simplified example smart contract code to illustrate data-flows between functions.</p>
Full article ">Figure 2
<p>Two state variables packed in the single slot of the EVM storage.</p>
Full article ">Figure 3
<p>Comparison of the number of bugs found over time.</p>
Full article ">
34 pages, 42799 KiB  
Article
YOLO-DentSeg: A Lightweight Real-Time Model for Accurate Detection and Segmentation of Oral Diseases in Panoramic Radiographs
by Yue Hua, Rui Chen and Hang Qin
Electronics 2025, 14(4), 805; https://doi.org/10.3390/electronics14040805 - 19 Feb 2025
Abstract
Panoramic radiography is vital in dentistry, where accurate detection and segmentation of diseased regions aid clinicians in fast, precise diagnosis. However, the current methods struggle with accuracy, speed, feature extraction, and suitability for low-resource devices. To overcome these challenges, this research introduces a [...] Read more.
Panoramic radiography is vital in dentistry, where accurate detection and segmentation of diseased regions aid clinicians in fast, precise diagnosis. However, the current methods struggle with accuracy, speed, feature extraction, and suitability for low-resource devices. To overcome these challenges, this research introduces a unique YOLO-DentSeg model, a lightweight architecture designed for real-time detection and segmentation of oral dental diseases, which is based on an enhanced version of the YOLOv8n-seg framework. First, the C2f(Channel to Feature Map)-Faster structure is introduced in the backbone network, achieving a lightweight design while improving the model accuracy. Next, the BiFPN(Bidirectional Feature Pyramid Network) structure is employed to enhance its multi-scale feature extraction capabilities. Then, the EMCA(Enhanced Efficient Multi-Channel Attention) attention mechanism is introduced to improve the model’s focus on key disease features. Finally, the Powerful-IOU(Intersection over Union) loss function is used to optimize the detection box localization accuracy. Experiments show that YOLO-DentSeg achieves a detection precision (mAP50(Box)) of 87%, segmentation precision (mAP50(Seg)) of 85.5%, and a speed of 90.3 FPS. Compared to YOLOv8n-seg, it achieves superior precise and faster inference times while decreasing the model size, computational load, and parameter count by 44.9%, 17.5%, and 44.5%, respectively. YOLO-DentSeg enables fast, accurate disease detection and segmentation, making it practical for devices with limited computing power and ideal for real-world dental applications. Full article
Show Figures

Figure 1

Figure 1
<p>Schematic diagram of oral disease detection and segmentation.</p>
Full article ">Figure 2
<p>YOLO-DentSeg model structure diagram.</p>
Full article ">Figure 3
<p>PConv schematic.</p>
Full article ">Figure 4
<p>Comparison of Faster-Block and Bottleneck structures.</p>
Full article ">Figure 5
<p>C2f-Faster schematic.</p>
Full article ">Figure 6
<p>FPN, PANet, and BiFPN structures.</p>
Full article ">Figure 7
<p>EMCA schematic of attention mechanisms.</p>
Full article ">Figure 8
<p>Schematics of CIOU and PowerIOU. (<b>a</b>) The structure of the original YOLOv8 boundary box loss function, CIoU (Complete Intersection over Union); (<b>b</b>) The structure of the proposed boundary box loss function, Powerful-IoU.</p>
Full article ">Figure 9
<p>The images before and after data augmentation.</p>
Full article ">Figure 10
<p>Comparison of detection and segmentation accuracy averages prior to and following model enhancement.</p>
Full article ">Figure 11
<p>Experimental curves for ablation experiments.</p>
Full article ">Figure 12
<p>Adding experimental curves for different attention modules.</p>
Full article ">Figure 13
<p>Experimental curves with various employed loss functions.</p>
Full article ">Figure 14
<p>Scatterplots of different model experiments. (<b>A</b>) The relationship between the number of parameters and FPS (Frames Per Second) for each model; (<b>B</b>) The relationship between computational complexity (FLOPs) and FPS for each model; (<b>C</b>) The relationship between FPS and mAP50 (Box) for each model; (<b>D</b>) The relationship between FPS and mAP50 (Seg) for each model.</p>
Full article ">Figure 15
<p>Detection segmentation results for different models.</p>
Full article ">
17 pages, 2073 KiB  
Article
Few-Shot Learning with Multimodal Fusion for Efficient Cloud–Edge Collaborative Communication
by Bo Gao, Xing Liu and Quan Zhou
Electronics 2025, 14(4), 804; https://doi.org/10.3390/electronics14040804 - 19 Feb 2025
Abstract
As demand for high-capacity, low-latency communication rises, mmWave systems are essential for enabling ultra-high-speed transmission in fifth-generation mobile communication technology (5G) and upcoming 6G networks, especially in dynamic, data-scarce environments. However, deploying mmWave systems in dynamic environments presents significant challenges, especially in beam [...] Read more.
As demand for high-capacity, low-latency communication rises, mmWave systems are essential for enabling ultra-high-speed transmission in fifth-generation mobile communication technology (5G) and upcoming 6G networks, especially in dynamic, data-scarce environments. However, deploying mmWave systems in dynamic environments presents significant challenges, especially in beam selection, where limited training data and environmental variability hinder optimal performance. In such scenarios, computation offloading has emerged as a key enabler, allowing computationally intensive tasks to be shifted from resource-constrained edge devices to powerful cloud servers, thereby reducing latency and optimizing resource utilization. This paper introduces a novel cloud–edge collaborative approach integrating few-shot learning (FSL) with multimodal fusion to address these challenges. By leveraging data from diverse modalities—such as red-green-blue (RGB) images, radar signals, and light detection and ranging (LiDAR)—within a cloud–edge architecture, the proposed framework effectively captures spatiotemporal features, enabling efficient and accurate beam selection with minimal data requirements. The cloud server is tasked with computationally intensive training, while the edge node focuses on real-time inference, ensuring low-latency decision making. Experimental evaluations confirm the model’s robustness, achieving high beam selection accuracy under one-shot and five-shot conditions while reducing computational overhead. This study highlights the potential of combining cloud–edge collaboration with FSL and multimodal fusion for next-generation wireless networks, paving the way for scalable, intelligent, and adaptive mmWave communication systems. Full article
(This article belongs to the Special Issue Computation Offloading for Mobile-Edge/Fog Computing)
Show Figures

Figure 1

Figure 1
<p>The proposed beam selection for cloud–edge collaboration in MIMO communication systems.</p>
Full article ">Figure 2
<p>Illustration of the proposed few-shot beam prediction model and its components. (<b>a</b>) The feature extraction module. (<b>b</b>) The proposed beam prediction model.</p>
Full article ">Figure 3
<p>The accuracy of the proposed model under 1-shot and 5-shot conditions fluctuates with a training-to-test-set ratio of 1:1.</p>
Full article ">Figure 4
<p>Accuracies of different Transformer modules (multi-attention layers) under 1-shot conditions.</p>
Full article ">Figure 5
<p>The impact of the number of CNN layers on the results.</p>
Full article ">Figure 6
<p>Accuracy of different models and multimodal fusion under 1-shot conditions.</p>
Full article ">Figure 7
<p>Accuracies of the proposed approach and various baseline methods.</p>
Full article ">Figure 8
<p>The proposed algorithm was evaluated against the baseline algorithms in terms of inference time for a single test data point following the training phase.</p>
Full article ">
16 pages, 2894 KiB  
Article
Frequency Multipliers Based on a Dual-Gate Graphene FET with M-Shaped Resistance Characteristics on a Flexible Substrate
by Jiaojiao Tian, Pei Peng, Zhongyang Ren, Chenhao Xia, Liming Ren, Fei Liu and Yunyi Fu
Electronics 2025, 14(4), 803; https://doi.org/10.3390/electronics14040803 - 19 Feb 2025
Abstract
Frequency multipliers are essential components in communication systems, and graphene’s exceptional electrical properties make it highly promising for flexible electronics. This paper addresses the technical challenges of multi-frequency multipliers based on graphene field-effect transistors (GFETs) and introduces a novel fabrication method using graphene [...] Read more.
Frequency multipliers are essential components in communication systems, and graphene’s exceptional electrical properties make it highly promising for flexible electronics. This paper addresses the technical challenges of multi-frequency multipliers based on graphene field-effect transistors (GFETs) and introduces a novel fabrication method using graphene as the channel material and metals with different work functions as the top gate. By employing Ti and Pd with distinct work functions, we develop a dual-gate GFET device that exhibits stable M-shaped resistance characteristics on a flexible polyethylene naphthalate (PEN) substrate. We demonstrate frequency doubler, tripler, and quadrupler on the flexible substrate. The results show that the GFET-based frequency multiplier offers advantages such as low operating voltage (<1 V), high voltage conversion efficiency (up to 8.4% for tripler and 6% for quadrupler), and high spectral purity (up to 88% for tripler and 76% for quadrupler). The intrinsic maximum operating frequency of the frequency quadrupler reaches 54 GHz. The use of a monolayer graphene channel, dual-metal gate control enabling an M-shaped transfer curve, and flexible characteristics all contribute to its superior performance compared to conventional devices. Full article
Show Figures

Figure 1

Figure 1
<p>Schematic diagram of the process for fabricating top-gate and dual top-gate graphene field-effect transistors on a flexible substrate. (<b>a</b>) Transfer and patterning of graphene onto a flexible polyethylene naphthalate (PEN) substrate. (<b>b</b>) Fabrication of source and drain electrodes. (<b>c</b>) Deposition of the gate dielectric layer. (<b>d</b>) Evaporation of the top gate metal. Process of preparing dual top-gate graphene field-effect transistors on a flexible substrate. (<b>e</b>) Preparation of the Ti/Au top-gate electrode. (<b>f</b>) Preparation of the Pd/Au top-gate electrode. (<b>g</b>) Optical microscope image of the fabricated flexible graphene field-effect transistor. (<b>h</b>) Optical microscope images of the fabricated bimetallic gate graphene field-effect transistors (GFETs).</p>
Full article ">Figure 2
<p>Schematic of two GFETs with Λ-shaped resistance-gate voltage (R-V<sub>g</sub>) curves connected in series to form an M-shaped electrical characteristic curve. (<b>a</b>) Equivalent circuit diagram of series-connected GFETs, where R<sub>1</sub> and R<sub>2</sub> represent the resistances of the two GFETs with separated Dirac point voltages and R is their total series resistance. (<b>b</b>) Two GFETs with separated Dirac points connected in series to create an M-shaped R-V<sub>g</sub> curve. (<b>c</b>) Insufficient separation between Dirac points of series-connected GFETs. (<b>d</b>) One of the resistance peaks in the series-connected GFETs is too small.</p>
Full article ">Figure 3
<p>GFETs with metal top gates of varying work functions on a flexible substrate. (<b>a</b>) Structure of the metal gate GFET. (<b>b</b>) Work function distribution for graphene, gate oxide, Ti, and Pd. (<b>c</b>) R-V<sub>g</sub> curves comparison of Ti and Pd top gate metals (gate length: 3 μm, width: 5 μm, V<sub>d</sub>: 0.3 V). (<b>d</b>) Dirac point voltage statistics for GFETs with Ti and Pd gate metals.</p>
Full article ">Figure 4
<p>Dual-gate flexible GFETs and their electrical characteristics. (<b>a</b>) Simplified structure of the dual-gate GFET, with L<sub>Ti</sub> and L<sub>Pd</sub> representing the lengths of the Ti and Pd top grids. (<b>b</b>) Full structure of the dual-gate GFETs. (<b>c</b>) Optical microscope image of a typical device. (<b>d</b>,<b>f</b>,<b>h</b>) AFM images of three devices with different L<sub>Ti</sub> to L<sub>Pd</sub> ratios (scale: 3 μm), showing height profiles along the red line. (<b>e</b>,<b>g</b>,<b>i</b>) Resistance curves of devices with varying L<sub>Ti</sub> to L<sub>Pd</sub> ratios, where D, S, and G represent the drain, source, and gate electrodes, respectively.</p>
Full article ">Figure 5
<p>Effect of flexible substrate bending on the electrical characteristics of dual-gate GFETs. (<b>a</b>) Photos of flexible substrate with GFET array attached to a semi-cylindrical surface for electrical testing. (<b>b</b>) Schematic of device bending, with r representing the bending radius and M1 and M2 indicating the two top-gate metal layers. (<b>c</b>) Effect of bending radius on the electrical characteristics (R-V<sub>g</sub> curve, V<sub>d</sub> = 0.3 V). (<b>d</b>) Effect of bending cycles (N) on the electrical characteristics of GFETs (R-V<sub>g</sub> curve V<sub>d</sub> = 0.1 V).</p>
Full article ">Figure 6
<p>High-performance frequency multipliers using dual-gate flexible GFETs. (<b>a</b>) Schematic of the test circuit. (<b>b</b>) R-V<sub>g</sub> curve of the selected GFET device. The blue, yellow, and green boxes respectively show one operating regions for the second, third, and fourth harmonics. (<b>c</b>) Input/output voltage curves and frequency spectrum of the doubler. (<b>d</b>) Input/output voltage curves and frequency spectrum of the tripler. (<b>e</b>) Input/output voltage curves and frequency spectrum of the quadrupler.</p>
Full article ">Figure 7
<p>Frequency response of multiple frequency multipliers limited by cable capacitance. (<b>a</b>) Output signal variation with input frequency in quadrupler mode. (<b>b</b>) Equivalent circuit. (<b>c</b>) Comparison of amplitude-frequency response (black dots) and simulation results (solid line) for the quadruple frequency output signal.</p>
Full article ">
15 pages, 1589 KiB  
Article
End-to-End Speech Recognition with Deep Fusion: Leveraging External Language Models for Low-Resource Scenarios
by Lusheng Zhang, Shie Wu and Zhongxun Wang
Electronics 2025, 14(4), 802; https://doi.org/10.3390/electronics14040802 - 19 Feb 2025
Abstract
With the rapid development of Automatic Speech Recognition (ASR) technology, end-to-end speech recognition systems have gained significant attention due to their ability to directly convert raw speech signals into text. However, such systems heavily rely on large amounts of labeled speech data, which [...] Read more.
With the rapid development of Automatic Speech Recognition (ASR) technology, end-to-end speech recognition systems have gained significant attention due to their ability to directly convert raw speech signals into text. However, such systems heavily rely on large amounts of labeled speech data, which severely limits model training performance and generalization, especially in low-resource language environments. To address this issue, this paper proposes an end-to-end speech recognition approach based on deep fusion, which tightly integrates an external language model (LM) with the end-to-end model during the training phase, effectively compensating for the lack of linguistic prior knowledge. Unlike traditional shallow fusion methods, deep fusion enables the model and the external LM to share representations and jointly optimize during training, thereby enhancing recognition performance under low-resource conditions. Experiments conducted on the Common Voice dataset show that, in a 10 h extremely low-resource scenario, the deep fusion method reduces the character error rate (CER) from 51.1% to 17.65%. In a 100 h scenario, it achieves a relative reduction of approximately 2.8%. Furthermore, ablation studies on model layers demonstrate that even with a reduced number of encoder and decoder layers to decrease model complexity, deep fusion continues to effectively leverage external linguistic priors, significantly improving performance in low-resource speech recognition tasks. Full article
Show Figures

Figure 1

Figure 1
<p>Zipformer architecture diagram.</p>
Full article ">Figure 2
<p>RNN-T architecture diagram.</p>
Full article ">Figure 3
<p>LM and RNN-T fusion architecture diagram.</p>
Full article ">Figure 4
<p>Flowchart.</p>
Full article ">
17 pages, 3547 KiB  
Article
Optimization of Passive Damping for LCL-Filtered AC Grid-Connected PV-Storage Integrated Systems
by Yue Zhang, Chenchen Song, Tao Wang and Kai Wang
Electronics 2025, 14(4), 801; https://doi.org/10.3390/electronics14040801 - 19 Feb 2025
Abstract
This paper conducts an in-depth study on the application of inductor-capacitor-inductor (LCL) filters in grid-connected photovoltaic (PV) inverters. First, the resonance issues associated with LCL filters are analyzed, and solutions are discussed, with a focus on the implementation of passive damping strategies. Various [...] Read more.
This paper conducts an in-depth study on the application of inductor-capacitor-inductor (LCL) filters in grid-connected photovoltaic (PV) inverters. First, the resonance issues associated with LCL filters are analyzed, and solutions are discussed, with a focus on the implementation of passive damping strategies. Various passive damping schemes, based on the placement of resistors (R), are compared and analyzed, ultimately selecting the capacitor branch series resistor as the optimal solution. During the design process, multiple parameters, such as total inductance, inverter-side inductance, grid-side inductance, capacitance, and damping resistors, are considered in light of their mutual constraints. Detailed analysis and optimization of these parameters are performed based on steady-state operation, current ripple, and power loss limitations. Finally, it is concluded that the passive damping solution using a series resistor in the capacitor branch meets the requirements for stable operation and efficient filtering. The optimal solutions are identified as R1 = 0, R2 = ∞, R3 ≠ 0, and R4 = ∞, providing a reliable and effective filtering solution for grid-connected PV inverter systems. Full article
(This article belongs to the Special Issue Technology and Approaches of Battery Energy Storage System)
Show Figures

Figure 1

Figure 1
<p>L filter topology in a PV-storage system including a three-phase or single-phase schematic.</p>
Full article ">Figure 2
<p>Bode plots of L-type filtering at different given inductance values.</p>
Full article ">Figure 3
<p>LC filter topology in a PV-storage system including a three-phase or single-phase schematic.</p>
Full article ">Figure 4
<p>LCL filter topology in a PV-storage system including a three-phase or single-phase schematic.</p>
Full article ">Figure 5
<p>Comparison of Bode plots of L and LCL filter transfer functions.</p>
Full article ">Figure 6
<p>Steady-state operation vector diagram of inverter AC side. (<b>a</b>) Pure inductive mode with the strictest inductance constraints. (<b>b</b>) Pure capacitive mode with relaxed inductance constraints. (<b>c</b>) General inverter operation showing voltage and current relationships. (<b>d</b>) Vector epresentation of grid-side voltage under steady-state conditions.</p>
Full article ">Figure 7
<p>LCL filter passive damping scheme topology.</p>
Full article ">Figure 8
<p>Bode plot of grid-side inductor series resistance transfer function. Bode plot of the transfer function for the grid-side inductor series resistance.</p>
Full article ">Figure 9
<p>Bode plot of grid-side inductor parallel resistance transfer function.</p>
Full article ">Figure 10
<p>Bode plot of capacitor branch series resistance transfer function.</p>
Full article ">Figure 11
<p>Bode plot of transfer function of capacitor branch parallel resistor.</p>
Full article ">
18 pages, 3409 KiB  
Review
Trends and Applications of Artificial Intelligence in Project Management
by Diego Vergara, Antonio del Bosque, Georgios Lampropoulos and Pablo Fernández-Arias
Electronics 2025, 14(4), 800; https://doi.org/10.3390/electronics14040800 - 19 Feb 2025
Abstract
The integration of artificial intelligence (AI) into project management (PM) transforms how projects are planned, executed, and monitored. The main objective of this study is to provide a comprehensive bibliometric analysis exploring trends, thematic areas, and future directions in AI applications in project [...] Read more.
The integration of artificial intelligence (AI) into project management (PM) transforms how projects are planned, executed, and monitored. The main objective of this study is to provide a comprehensive bibliometric analysis exploring trends, thematic areas, and future directions in AI applications in project management by examining publications from the last decade. This research uncovers dominant themes such as machine learning, decision making, information management, and resource optimization. The findings highlight the growing use of AI to enhance efficiency, accuracy, and innovation in PM processes, with recent trends favoring data-driven approaches and emerging technologies like generative AI. Geographically, China, India, and the United States lead in publications, while the United Kingdom and Australia show a high citation impact. The research landscape, including AI-enhanced decision-making frameworks and cost analysis, demonstrates the diversity of AI applications in PM. An increased interest in the use of generative AI and its impact on PM and project managers was observed. This analysis contributes to the field by offering a structured overview of research trends, defining the challenges and opportunities for integrating AI into project management practices and offering perspectives on emerging technologies. Full article
Show Figures

Figure 1

Figure 1
<p>AI applications in project management.</p>
Full article ">Figure 2
<p>Identification of studies via PRISMA 2020 protocol.</p>
Full article ">Figure 3
<p>Document collection.</p>
Full article ">Figure 4
<p>Distribution of annual publications.</p>
Full article ">Figure 5
<p>Countries’ collaboration.</p>
Full article ">Figure 6
<p>Most common Keywords Plus.</p>
Full article ">Figure 7
<p>Most common Author Keywords.</p>
Full article ">Figure 8
<p>Keyword co-occurrence network.</p>
Full article ">Figure 9
<p>Trend topics.</p>
Full article ">Figure 10
<p>Document clusters.</p>
Full article ">Figure 11
<p>Thematic map.</p>
Full article ">
33 pages, 2092 KiB  
Article
SentimentFormer: A Transformer-Based Multimodal Fusion Framework for Enhanced Sentiment Analysis of Memes in Under-Resourced Bangla Language
by Fatema Tuj Johora Faria, Laith H. Baniata, Mohammad H. Baniata, Mohannad A. Khair, Ahmed Ibrahim Bani Ata, Chayut Bunterngchit and Sangwoo Kang
Electronics 2025, 14(4), 799; https://doi.org/10.3390/electronics14040799 (registering DOI) - 18 Feb 2025
Abstract
Social media has increasingly relied on memes as a tool for expressing opinions, making meme sentiment analysis an emerging area of interest for researchers. While much of the research has focused on English-language memes, under-resourced languages, such as Bengali, have received limited attention. [...] Read more.
Social media has increasingly relied on memes as a tool for expressing opinions, making meme sentiment analysis an emerging area of interest for researchers. While much of the research has focused on English-language memes, under-resourced languages, such as Bengali, have received limited attention. Given the surge in social media use, the need for sentiment analysis of memes in these languages has become critical. One of the primary challenges in this field is the lack of benchmark datasets, particularly in languages with fewer resources. To address this, we used the MemoSen dataset, designed for Bengali, which consists of 4368 memes annotated with three sentiment labels: positive, negative, and neutral. MemoSen is divided into training (70%), test (20%), and validation (10%) sets, with an imbalanced class distribution: 1349 memes in the positive class, 2728 in the negative class, and 291 in the neutral class. Our approach leverages advanced deep learning techniques for multimodal sentiment analysis in Bengali, introducing three hybrid approaches. SentimentTextFormer is a text-based, fine-tuned model that utilizes state-of-the-art transformer architectures to accurately extract sentiment-related insights from Bengali text, capturing nuanced linguistic features. SentimentImageFormer is an image-based model that employs cutting-edge transformer-based techniques for precise sentiment classification through visual data. Lastly, SentimentFormer is a hybrid model that seamlessly integrates both text and image modalities using fusion strategies. Early fusion combines textual and visual features at the input level, enabling the model to jointly learn from both modalities. Late fusion merges the outputs of separate text and image models, preserving their individual strengths for the final prediction. Intermediate fusion integrates textual and visual features at intermediate layers, refining their interactions during processing. These fusion strategies combine the strengths of both textual and visual data, enhancing sentiment analysis by exploiting complementary information from multiple sources. The performance of our models was evaluated using various accuracy metrics, with SentimentTextFormer achieving 73.31% accuracy and SentimentImageFormer attaining 64.72%. The hybrid model, SentimentFormer (SwiftFormer with mBERT), employing intermediate fusion, shows a notable improvement in accuracy, achieving 79.04%, outperforming SentimentTextFormer by 5.73% and SentimentImageFormer by 14.32%. Among the fusion strategies, SentimentFormer (SwiftFormer with mBERT) achieved the highest accuracy of 79.04%, highlighting the effectiveness of our fusion technique and the reliability of our multimodal framework in improving sentiment analysis accuracy across diverse modalities. Full article
Show Figures

Figure 1

Figure 1
<p>Distribution of samples Across train, test, and validation sets in the MemoSen dataset.</p>
Full article ">Figure 2
<p>Representative examples from the MemoSen dataset, illustrating memes labeled with positive, neutral, and negative sentiments.</p>
Full article ">Figure 3
<p>Unimodal sentiment classification framework for Bangla meme captions.</p>
Full article ">Figure 4
<p>Unimodal sentiment classification framework for meme images.</p>
Full article ">Figure 5
<p>Fusion framework for enhanced multimodal sentiment analysis of Bangla memes.</p>
Full article ">Figure 6
<p>Confusion matrices of SentimentTextFormer, SentimentImageFormer, and SentimentFormer showcasing their sentiment classification performance on the MemoSen dataset.</p>
Full article ">Figure 7
<p>Error analysis of multimodal sentiment classification in Bengali memes.</p>
Full article ">
15 pages, 3831 KiB  
Article
Narrowband Interference Cancellation Using a Fine Frequency Shift in Single-Carrier Frequency Domain Equalization (SC-FDE) Systems
by Seung-Mi Yun, Yeong-Bin Ryu, Dong-Ho Song and Eui-Rim Jeong
Electronics 2025, 14(4), 798; https://doi.org/10.3390/electronics14040798 - 18 Feb 2025
Abstract
This paper proposes a narrowband interference (NBI) cancellation technique for single-carrier frequency domain equalization (SC-FDE) systems. SC-FDE is a transmission scheme commonly used in mobile communication uplinks or low-Earth-orbit (LEO) satellite communications due to its low peak-to-average power ratio (PAPR) characteristics. In the [...] Read more.
This paper proposes a narrowband interference (NBI) cancellation technique for single-carrier frequency domain equalization (SC-FDE) systems. SC-FDE is a transmission scheme commonly used in mobile communication uplinks or low-Earth-orbit (LEO) satellite communications due to its low peak-to-average power ratio (PAPR) characteristics. In the presence of narrowband interference, removing the interference is crucial, as SC-FDE systems are vulnerable to the interference. A straightforward approach to interference removal is frequency nulling, which can be easily implemented using FFT (Fast Fourier Transform). However, this method is only effective when the interference frequency coincides with specific FFT grid frequencies. To address this limitation, this paper proposes identifying the interference frequency, applying a fine frequency shift, and then canceling or nulling the interference. After that, the fine frequency is shifted back for reception. By aligning the interference frequency with an FFT grid frequency, the proposed technique enables simple and effective narrowband interference cancellation. The performance of the proposed method is validated through computer simulations, which demonstrate excellent interference cancellation performance regardless of the signal-to-interference ratio (SIR). Full article
(This article belongs to the Section Microwave and Wireless Communications)
Show Figures

Figure 1

Figure 1
<p>SC-FDE structure and equalization.</p>
Full article ">Figure 2
<p>Block diagram of interference cancellation.</p>
Full article ">Figure 3
<p>Spectrum before cancellation (<b>left</b>) and after cancellation (<b>right</b>) when <math display="inline"><semantics> <mrow> <msub> <mrow> <mi mathvariant="bold-italic">f</mi> </mrow> <mrow> <mi mathvariant="bold-italic">I</mi> </mrow> </msub> <mo>=</mo> <mn>50</mn> <mo>/</mo> <mn>256</mn> <mi mathvariant="bold-italic">T</mi> </mrow> </semantics></math>.</p>
Full article ">Figure 4
<p>Spectrum before cancellation (<b>left</b>) and after cancellation (<b>right</b>) when <math display="inline"><semantics> <mrow> <msub> <mrow> <mi mathvariant="bold-italic">f</mi> </mrow> <mrow> <mi mathvariant="bold-italic">I</mi> </mrow> </msub> <mo>=</mo> <mn>50.4</mn> <mo>/</mo> <mn>256</mn> <mi mathvariant="bold-italic">T</mi> </mrow> </semantics></math>.</p>
Full article ">Figure 5
<p>Proposed fine frequency shift.</p>
Full article ">Figure 6
<p>Spectrum before cancellation (<b>left</b>) and after cancellation (<b>right</b>) with proposed fine frequency shift when <math display="inline"><semantics> <mrow> <msub> <mrow> <mi mathvariant="bold-italic">f</mi> </mrow> <mrow> <mi mathvariant="bold-italic">I</mi> </mrow> </msub> <mo>=</mo> <mn>50.4</mn> <mo>/</mo> <mn>256</mn> <mi mathvariant="bold-italic">T</mi> </mrow> </semantics></math>.</p>
Full article ">Figure 7
<p>Transceiver model for simulation.</p>
Full article ">Figure 8
<p>RMSE performance for fine frequency estimation under SIR = 5 dB (<b>left</b>): AWGN channel; (<b>right</b>): ETU channel.</p>
Full article ">Figure 9
<p>RMSE performance for fine frequency estimation under SIR = 0 dB (<b>left</b>): AWGN channel; (<b>right</b>): ETU channel.</p>
Full article ">Figure 10
<p>BER performance with interference cancellation under the AWGN channel and SIR = 5 dB (solid: spreading gain 1, dashed: spreading gain 4).</p>
Full article ">Figure 11
<p>BER performance with interference cancellation under AWGN channel and SIR = 0 dB (solid: spreading gain 1, dashed: spreading gain 4).</p>
Full article ">Figure 12
<p>BER performance with interference cancellation under ETU channel and SIR = 5 dB (solid: spreading gain 1, dashed: spreading gain 4).</p>
Full article ">Figure 13
<p>BER performance with interference cancellation under ETU channel and SIR = 0 dB (solid: spreading gain 1, dashed: spreading gain 4).</p>
Full article ">
27 pages, 15796 KiB  
Article
MSFF: A Multi-Scale Feature Fusion Convolutional Neural Network for Hyperspectral Image Classification
by Gu Gong, Xiaopeng Wang, Jiahua Zhang, Xiaodi Shang, Zhicheng Pan, Zhiyuan Li and Junshi Zhang
Electronics 2025, 14(4), 797; https://doi.org/10.3390/electronics14040797 - 18 Feb 2025
Abstract
In contrast to conventional remote sensing images, hyperspectral remote sensing images are characterized by a greater number of spectral bands and exceptionally high resolution. The richness of both spectral and spatial information facilitates the precise classification of various objects within the images, establishing [...] Read more.
In contrast to conventional remote sensing images, hyperspectral remote sensing images are characterized by a greater number of spectral bands and exceptionally high resolution. The richness of both spectral and spatial information facilitates the precise classification of various objects within the images, establishing hyperspectral imaging as indispensable for remote sensing applications. However, the labor-intensive and time-consuming process of labeling hyperspectral images results in limited labeled samples, while challenges like spectral similarity between different objects and spectral variation within the same object further complicate the development of classification algorithms. Therefore, efficiently exploiting the spatial and spectral information in hyperspectral images is crucial for accomplishing the classification task. To address these challenges, this paper presents a multi-scale feature fusion convolutional neural network (MSFF). The network introduces a dual branch spectral and spatial feature extraction module utilizing 3D depthwise separable convolution for joint spectral and spatial feature extraction, further refined by an attention-based-on-central-pixels (ACP) mechanism. Additionally, a spectral–spatial joint attention module (SSJA) is designed to interactively explore latent dependency between spectral and spatial information through the use of multilayer perceptron and global pooling operations. Finally, a feature fusion module (FF) and an adaptive multi-scale feature extraction module (AMSFE) are incorporated to enable adaptive feature fusion and comprehensive mining of feature information. Experimental results demonstrate that the proposed method performs exceptionally well on the IP, PU, and YRE datasets, delivering superior classification results compared to other methods and underscoring the potential and advantages of MSFF in hyperspectral remote sensing classification. Full article
(This article belongs to the Special Issue Machine Learning and Computational Intelligence in Remote Sensing)
Show Figures

Figure 1

Figure 1
<p>Overall architecture of the MSFF model.</p>
Full article ">Figure 2
<p>Detailed architecture of the MSFF model.</p>
Full article ">Figure 3
<p>3D convolution.</p>
Full article ">Figure 4
<p>3D depthwise separable convolution.</p>
Full article ">Figure 5
<p>The proposed ACP structure.</p>
Full article ">Figure 6
<p>Structure of SSFE module.</p>
Full article ">Figure 7
<p>The proposed SSJA structure.</p>
Full article ">Figure 8
<p>The proposed FF structure.</p>
Full article ">Figure 9
<p>The proposed NAM structure.</p>
Full article ">Figure 10
<p>The proposed AMSFE structure.</p>
Full article ">Figure 11
<p>Datasets and ground truth: (<b>1</b>,<b>2</b>) Indian Pines dataset, (<b>3</b>,<b>4</b>) Pavia University dataset, and (<b>5</b>,<b>6</b>) Yellow River Estuary coastal wetland.</p>
Full article ">Figure 12
<p>Results of different batch sizes on OA, AA, and Kappa.</p>
Full article ">Figure 13
<p>Results of different patch sizes on OA, AA, and Kappa.</p>
Full article ">Figure 14
<p>Results of different channel numbers on OA, AA, and Kappa.</p>
Full article ">Figure 15
<p>Results of different FEG numbers on OA, AA, and Kappa.</p>
Full article ">Figure 16
<p>Classification maps for the IP dataset. (<b>a</b>) Ground-truth map. (<b>b</b>) 2D-CNN. (<b>c</b>) DBMA. (<b>d</b>) SSRN. (<b>e</b>) HybridSN. (<b>f</b>) MDRD-Net. (<b>g</b>) SSFTT. (<b>h</b>) DMAN. (<b>i</b>) MSFF-CBAM. (<b>j</b>) MSFF.</p>
Full article ">Figure 17
<p>Classification maps for the PU dataset. (<b>a</b>) Ground-truth map. (<b>b</b>) 2D-CNN. (<b>c</b>) DBMA. (<b>d</b>) SSRN. (<b>e</b>) HybridSN. (<b>f</b>) MDRD-Net. (<b>g</b>) SSFTT. (<b>h</b>) DMAN. (<b>i</b>) MSFF-CBAM. (<b>j</b>) MSFF.</p>
Full article ">Figure 18
<p>Classification maps for the YRE dataset. (<b>a</b>) Ground-truth map. (<b>b</b>) 2D-CNN. (<b>c</b>) DBMA. (<b>d</b>) SSRN. (<b>e</b>) HybridSN. (<b>f</b>) MDRD-Net. (<b>g</b>) SSFTT. (<b>h</b>) DMAN. (<b>i</b>) MSFF-CBAM. (<b>j</b>) MSFF.</p>
Full article ">Figure 19
<p>The OA of various combinations in three datasets.</p>
Full article ">
22 pages, 4139 KiB  
Article
Estimation of Uncertain Parameters in Single and Double Diode Models of Photovoltaic Panels Using Frilled Lizard Optimization
by Süleyman Dal and Necmettin Sezgin
Electronics 2025, 14(4), 796; https://doi.org/10.3390/electronics14040796 - 18 Feb 2025
Abstract
Renewable energy sources are increasingly crucial for sustainable development. Photovoltaic (PV) systems, which convert solar energy into electricity, offer an environmentally friendly solution. Enhancing energy efficiency and minimizing environmental impacts in these systems heavily rely on parameter optimization. In this study, the Frilled [...] Read more.
Renewable energy sources are increasingly crucial for sustainable development. Photovoltaic (PV) systems, which convert solar energy into electricity, offer an environmentally friendly solution. Enhancing energy efficiency and minimizing environmental impacts in these systems heavily rely on parameter optimization. In this study, the Frilled Lizard Optimization (FLO) algorithm is proposed as a novel approach, integrating the newton-raphson method into the root mean square error (RMSE) objective function process to address nonlinear equations. Extensive analyses conducted on RTC France, STM6-40/36, and Photowatt PWP201 modules demonstrate the superior performance of the FLO algorithm using MATLAB R2022a software with Intel(R) Core(TM) i7-7500U CPU@ 2.70GHz 2.90 GHz 8 GB RAM. The RMSE values were calculated as 0.0030375 and 0.011538 for SDM and DDM in the RTC France dataset, 0.012036 for the STM6-40/36 dataset and 0.0097545 for the Photowatt-PWP201 dataset, respectively, indicating significantly lower error margins compared to other optimisation methods. Additionally, comprehensive evaluations were carried out using error metrics such as individual absolute error (IAE), relative error (RE) and mean absolute error (MAE), supported by detailed graphical representations of measured and predicted parameters. Current-voltage (I-V) and power-voltage (P-V) characteristic curves, as well as convergence behaviors, were systematically analyzed. This study introduces an innovative and robust solution for parameter optimization in PV systems, contributing to both theoretical and industrial applications. Full article
(This article belongs to the Special Issue Machine Learning Techniques for Image Processing)
Show Figures

Figure 1

Figure 1
<p>Equivalent circuit of photovoltaic models.</p>
Full article ">Figure 2
<p>An overview of FLO algorithm-based parameter estimation of PV models.</p>
Full article ">Figure 3
<p>FLO flowchart.</p>
Full article ">Figure 4
<p>Comparison between the measured and the estimated data obtained by FLO algorithm for R.T.C France model (<b>a</b>) I-V of SDM, (<b>b</b>) P-V of SDM.</p>
Full article ">Figure 5
<p>Comparison between the measured and the estimated data obtained by FLO algorithm for R.T.C France model (<b>a</b>) I-V of DDM, (<b>b</b>) P-V of DDM.</p>
Full article ">Figure 6
<p>Convergence graph of proposed FLO algorithm for R.T.C France model: (<b>a</b>) SDM, (<b>b</b>) DDM.</p>
Full article ">Figure 7
<p>Performance Comparison of Single and Double Diode Models Using FLO algorithm (<b>a</b>) I-V, (<b>b</b>) P-V and (<b>c</b>) Error Metrics for Each Data Point.</p>
Full article ">Figure 8
<p>Comparison between the measured and the estimated data obtained by FLO algorithm for STM6-40/36 model (<b>a</b>) I-V curve. (<b>b</b>) P-V curve.</p>
Full article ">Figure 9
<p>Comparison between the measured and the estimated data obtained by FLO algorithm for Photowatt PWP-201 model (<b>a</b>) I-V curve. (<b>b</b>) P-V curve.</p>
Full article ">Figure 10
<p>Convergence graph of proposed FLO algorithm for: (<b>a</b>) STM6-40/36. (<b>b</b>) Photowatt PWP-201.</p>
Full article ">Figure 11
<p>Performance Comparison I-V, P-V and Error Metrics for Each Data Point; (<b>a</b>,<b>c</b>) STM6-40/36 and (<b>b</b>,<b>d</b>) Photowatt-PWP201.</p>
Full article ">Figure 12
<p>Statistical scores (RMSE) of all runs (30) by FLO algorithm: (<b>a</b>) SDM. (<b>b</b>) DDM. (<b>c</b>) STM6-40/36 and (<b>d</b>) Photowatt PWP-201.</p>
Full article ">
15 pages, 9408 KiB  
Article
Graph Isomorphic Network-Assisted Optimal Coordination of Wave Energy Converters Based on Maximum Power Generation
by Ashkan Safari, Afshin Rahimi and Hoda Sorouri
Electronics 2025, 14(4), 795; https://doi.org/10.3390/electronics14040795 - 18 Feb 2025
Abstract
Oceans are a major source of clean energy, harnessing the vast and consistent power of waves to generate electricity. Today, they are seen as a vital renewable and clean solution for transitioning to a complete fossil fuel-free future world. To get the most [...] Read more.
Oceans are a major source of clean energy, harnessing the vast and consistent power of waves to generate electricity. Today, they are seen as a vital renewable and clean solution for transitioning to a complete fossil fuel-free future world. To get the most out of ocean wave potential, Wave Energy Converters (WECs) are being used to harness the power of ocean waves into usable electrical energy. To this end, to maximize the power generated from the WECs, two strategies for WEC design improvement and optimal coordination can be considered. Among these, optimal coordination is the more straightforward method to implement. However, most of the recently developed coordination strategies are dynamic-based, encountering challenges as the system’s scale expands and grows larger. Consequently, a novel Graph Isomorphic Network (GIN)-based model is presented in this paper. The proposed model consists of the following five layers: the input graph, two GIN convolutional layers (GIN Conv.1, and 2), a mean pooling layer, and the output layer. The target of total generated power is predicted based on the features of the generated power from each WEC and the related spatial coordinates {xi,yi}. Subsequently, based on the anticipated total power considered by the model, the system enables maximum generation. The model performs spatial coordination analyses to present the optimal coordination for each WEC to achieve the objective of maximizing total generated power. The proposed model is evaluated through several Key Performance Indicators (KPIs), achieving the least number of errors in prediction and optimal coordination performances. Full article
(This article belongs to the Special Issue Advances in Renewable Energy and Electricity Generation)
Show Figures

Figure 1

Figure 1
<p>Proposed system of 16 WECs, the data of which is considered for optimal coordination.</p>
Full article ">Figure 2
<p>Utilized GIN strategy in this work.</p>
Full article ">Figure 3
<p>The average power of each WEC, coordinated at various points.</p>
Full article ">Figure 4
<p>The overall contribution of each WEC, along with the variability present in their output power at various points.</p>
Full article ">Figure 5
<p>Detailed analytics of the output power range of each WEC, evaluated at various points. These analytics demonstrate the specific regions where each WEC exhibited a higher density of generated output power.</p>
Full article ">Figure 6
<p>Spatial coordination and power profile for WECs. Points near yellow indicate higher generated power.</p>
Full article ">Figure 7
<p>Density results of the real/predicted power by the proposed model. Density means distribution here, for instance, <math display="inline"><semantics> <mrow> <msub> <mi>P</mi> <mrow> <mi>P</mi> <mi>r</mi> <mi>e</mi> <mi>d</mi> </mrow> </msub> <mo>=</mo> <mn>1.47</mn> </mrow> </semantics></math> [MW] has a density of 10, which means there are 10 values of 1.47 [MW] in the predicted data.</p>
Full article ">Figure 8
<p>Optimal coordination results, derived by the model, for 16 WECs, based on <a href="#electronics-14-00795-t001" class="html-table">Table 1</a>.</p>
Full article ">
18 pages, 1942 KiB  
Article
Resume2Vec: Transforming Applicant Tracking Systems with Intelligent Resume Embeddings for Precise Candidate Matching
by Ravi Varma Kumar Bevara, Nishith Reddy Mannuru, Sai Pranathi Karedla, Brady Lund, Ting Xiao, Harshitha Pasem, Sri Chandra Dronavalli and Siddhanth Rupeshkumar
Electronics 2025, 14(4), 794; https://doi.org/10.3390/electronics14040794 - 18 Feb 2025
Abstract
Conventional Applicant Tracking Systems (ATSs) encounter considerable constraints in accurately aligning resumes with job descriptions (JD), especially in handling unstructured data and intricate qualifications. We provide Resume2Vec, an innovative method that utilizes transformer-based deep learning models, including encoders (BERT, RoBERTa, and DistilBERT) and [...] Read more.
Conventional Applicant Tracking Systems (ATSs) encounter considerable constraints in accurately aligning resumes with job descriptions (JD), especially in handling unstructured data and intricate qualifications. We provide Resume2Vec, an innovative method that utilizes transformer-based deep learning models, including encoders (BERT, RoBERTa, and DistilBERT) and decoders (GPT, Gemini, and Llama), to create embeddings for resumes and job descriptions, employing cosine similarity for evaluation. Our methodology integrates quantitative analysis via embedding-based evaluation with qualitative human assessment across several professional fields. Experimental findings indicate that Resume2Vec outperformed conventional ATS systems, achieving enhancements of up to 15.85% in Normalized Discounted Cumulative Gain (nDCG) and 15.94% in Ranked Biased Overlap (RBO) scores, especially within the mechanical engineering and health and fitness domains. Although conventional the ATS exhibited slightly superior nDCG scores in operations management and software testing, Resume2Vec consistently displayed a more robust alignment with human preferences across the majority of domains, as indicated by the RBO metrics. This research demonstrates that Resume2Vec is a powerful and scalable method for matching resumes to job descriptions, effectively overcoming the shortcomings of traditional systems, while preserving a high alignment with human evaluation criteria. The results indicate considerable promise for transformer-based methodologies in enhancing recruiting technology, facilitating more efficient and precise candidate selection procedures. Full article
(This article belongs to the Special Issue Big Data and AI Applications)
Show Figures

Figure 1

Figure 1
<p>Architecture of the proposed system for resume–JD mapping.</p>
Full article ">Figure 2
<p>Scatter plot of resume embeddings across domains.</p>
Full article ">Figure 3
<p>Scatter plot of job description embeddings across domains.</p>
Full article ">Figure 4
<p>Model accuracy comparison without PCA.</p>
Full article ">Figure 5
<p>Model accuracy comparison with PCA.</p>
Full article ">Figure 6
<p>Comparison of Resume2Vec and ATS performance across various metrics (nDCG and RBO) for different job categories.</p>
Full article ">
16 pages, 2664 KiB  
Article
Development of New Generation Portable Camera-Aided Surgical Simulator for Cognitive Training in Laparoscopic Cholecystectomy
by Yucheng Li, Victoria Nelson, Cuong T. Nguyen, Irene Suh, Suvranu De, Ka-Chun Siu and Carl Nelson
Electronics 2025, 14(4), 793; https://doi.org/10.3390/electronics14040793 - 18 Feb 2025
Abstract
Laparoscopic cholecystectomy (LC) is the standard procedure for gallbladder removal, but improper identification of anatomical structures can lead to biliary duct injury (BDI). The critical view of safety (CVS) is a standardized technique designed to mitigate this risk. However, existing surgical training systems [...] Read more.
Laparoscopic cholecystectomy (LC) is the standard procedure for gallbladder removal, but improper identification of anatomical structures can lead to biliary duct injury (BDI). The critical view of safety (CVS) is a standardized technique designed to mitigate this risk. However, existing surgical training systems primarily emphasize haptic feedback and physical skill development, making them expensive and less accessible. This paper presents the next-generation Portable Camera-Aided Surgical Simulator (PortCAS), a cost-effective, portable, vision-based surgical training simulator designed to enhance cognitive skill acquisition in LC. The system consists of an enclosed physical module equipped with a vision system, a single-board computer for real-time instrument tracking, and a virtual simulation interface that runs on a user-provided computer. Unlike traditional simulators, PortCAS prioritizes cognitive training over force-based interactions, eliminating the need for costly haptic components. The system was evaluated through user studies assessing accuracy, usability, and training effectiveness. Results demonstrate that PortCAS provides a sufficiently accurate tracking performance for training surgical skills such as CVS, offering a scalable and accessible solution for surgical education. Full article
(This article belongs to the Special Issue Virtual Reality Applications in Enhancing Human Lives)
Show Figures

Figure 1

Figure 1
<p>Workflow diagram of the portable surgical training simulator. The system comprises three components: (1) three smartphones equipped with cameras, an installed app, and an image segmentation program; (2) a Raspberry Pi running a triangulation program to estimate marker positions in space; and (3) a vision computer that renders the VR environment. Smartphones capture marker pixel coordinates and transmit the data to the Raspberry Pi. The Raspberry Pi processes the data to calculate marker positions and sends the results to the computer, which generates the immersive VR simulation.</p>
Full article ">Figure 2
<p>Portable enclosure design and assembly process. (<b>A</b>) Unfolded enclosure: The piece is laser-cut from <math display="inline"><semantics> <mrow> <mn>1</mn> <mo>/</mo> <msup> <mn>8</mn> <mrow> <mo>″</mo> </mrow> </msup> </mrow> </semantics></math> plywood and connected with <math display="inline"><semantics> <msup> <mn>12</mn> <mrow> <mo>″</mo> </mrow> </msup> </semantics></math> hinges using <math display="inline"><semantics> <mrow> <mn>1</mn> <mo>/</mo> <msup> <mn>8</mn> <mrow> <mo>″</mo> </mrow> </msup> </mrow> </semantics></math> rivets for foldability. (<b>B</b>) Folded enclosure: The compact design achieves a folded volume of <math display="inline"><semantics> <mrow> <mn>12</mn> <mo>.</mo> <msup> <mn>25</mn> <mrow> <mo>″</mo> </mrow> </msup> <mo>×</mo> <msup> <mn>12</mn> <mrow> <mo>″</mo> </mrow> </msup> <mo>×</mo> <mn>1</mn> <mo>.</mo> <msup> <mn>25</mn> <mrow> <mo>″</mo> </mrow> </msup> </mrow> </semantics></math> for portability. (<b>C</b>) Installed enclosure: Tabs and slots securely connect the panels to form the working structure. (<b>D</b>) Fully assembled prototype: The enclosure is equipped with three smartphones and two laparoscopic graspers, ready for simulation use.</p>
Full article ">Figure 3
<p>The schematic of the enclosure shows the enclosure design and layout for camera positioning and triangulation analysis. The Remote Center of Motion (RCM) is the fixed position where the laparoscopic gripper passes through and is secured within the enclosure. The red, green, and blue arrows represent the x-, y-, and z-axes, respectively.</p>
Full article ">Figure 4
<p>Color-based segmentation is applied to identify surgical instrument tips. (<b>A</b>) shows color markers, (<b>B</b>) highlights the segmented colors, and (<b>C</b>) shows centroids for triangulation.</p>
Full article ">Figure 5
<p>(<b>A</b>) Local view showing the rays from three cameras and the estimated target position. (<b>B</b>) Triangulation setup illustrating camera rays and target position. The red, green, and blue arrows represent the <span class="html-italic">x</span>-, <span class="html-italic">y</span>-, and <span class="html-italic">z</span>-axes, respectively.</p>
Full article ">Figure 6
<p>Test scenarios for assessing camera layouts. All camera configurations are positioned on the surface of a sphere centered on the target position. (<b>A</b>,<b>D</b>,<b>G</b>) correspond to <math display="inline"><semantics> <mrow> <mi>θ</mi> <mo>=</mo> <msup> <mn>80</mn> <mo>∘</mo> </msup> </mrow> </semantics></math>; (<b>B</b>,<b>E</b>,<b>H</b>) to <math display="inline"><semantics> <mrow> <mi>θ</mi> <mo>=</mo> <mn>54</mn> <mo>.</mo> <msup> <mn>8</mn> <mo>∘</mo> </msup> </mrow> </semantics></math>; and (<b>C</b>,<b>F</b>,<b>I</b>) to <math display="inline"><semantics> <mrow> <mi>θ</mi> <mo>=</mo> <msup> <mn>85</mn> <mo>∘</mo> </msup> </mrow> </semantics></math>. (<b>A</b>–<b>C</b>) have an azimuthal angle distribution of <math display="inline"><semantics> <mrow> <mi>ϕ</mi> <mo>=</mo> <msup> <mn>150</mn> <mo>∘</mo> </msup> </mrow> </semantics></math>; (<b>D</b>–<b>F</b>) have <math display="inline"><semantics> <mrow> <mi>ϕ</mi> <mo>=</mo> <msup> <mn>120</mn> <mo>∘</mo> </msup> </mrow> </semantics></math>; and (<b>G</b>–<b>I</b>) have <math display="inline"><semantics> <mrow> <mi>ϕ</mi> <mo>=</mo> <msup> <mn>60</mn> <mo>∘</mo> </msup> </mrow> </semantics></math>. The red, green, and blue arrows represent the <span class="html-italic">x</span>-, <span class="html-italic">y</span>-, and <span class="html-italic">z</span>-axes, respectively.</p>
Full article ">Figure 7
<p>Contour map of condition numbers for various camera layout scenarios, computed based on Equation (<a href="#FD9-electronics-14-00793" class="html-disp-formula">9</a>). The map covers a continuous range of <math display="inline"><semantics> <mi>θ</mi> </semantics></math> and <math display="inline"><semantics> <mi>ϕ</mi> </semantics></math>, illustrating the impact of these parameters on the condition number. The nine discrete scenarios (<b>A</b>–<b>I</b>) from <a href="#electronics-14-00793-f006" class="html-fig">Figure 6</a> are marked at their corresponding locations on the map for reference.</p>
Full article ">Figure 8
<p>(<b>A</b>) Illustration of the <math display="inline"><semantics> <mrow> <mn>5</mn> <mo>×</mo> <mn>5</mn> </mrow> </semantics></math> grid of target positions within the enclosure’s workspace, with each square measuring 15 mm <math display="inline"><semantics> <mrow> <mo>×</mo> <mn>15</mn> </mrow> </semantics></math> mm. (<b>B</b>) Positioning of the two symmetric cameras. (<b>C</b>) Positioning of the camera on the symmetric plane. (<b>D</b>) Schematic of the enclosure’s interior showing camera placement and target grid.</p>
Full article ">Figure 9
<p>Accuracy test results comparing estimated target positions (blue points) to ground truth target positions (red grid) across different planes: (<b>A</b>) <math display="inline"><semantics> <mrow> <mi>x</mi> <mi>y</mi> </mrow> </semantics></math>-plane, (<b>B</b>) <math display="inline"><semantics> <mrow> <mi>y</mi> <mi>z</mi> </mrow> </semantics></math>-plane, (<b>C</b>) <math display="inline"><semantics> <mrow> <mi>x</mi> <mi>z</mi> </mrow> </semantics></math>-plane, and (<b>D</b>) 3D view of the workspace. All dimensions are presented in millimeters. The red, green, and blue arrows represent the <span class="html-italic">x</span>-, <span class="html-italic">y</span>-, and <span class="html-italic">z</span>-axes, respectively.</p>
Full article ">Figure 10
<p>Simulation results of the VR environment for laparoscopic training. The system visualizes the gallbladder and surrounding organs with realistic coloration and texture to enhance realism. Interaction with virtual organs demonstrates key procedural steps, including connective tissue dissection and gallbladder isolation. (<b>A</b>) Illustration of the setup with two laparoscopic grippers. (<b>B</b>) Demonstration of both devices grasping either the liver or the gallbladder. (<b>C</b>) Illustration of the left arm grasping the liver while the right arm dissects fat tissue. (<b>D</b>) Depiction of the right arm grasping the liver while the left arm dissects fat tissue.</p>
Full article ">
Previous Issue
Next Issue
Back to TopTop