Electronics Science Technology and Application

The effectiveness of AI-based classroom activity design in college English instruction in China

Haiyan Wu; Azli Bin Ariffin — Thu, 02 Apr 2026 16:49:28 +0800

With the acceleration of educational digital transformation and the deepening of college English teaching reform, artificial intelligence (AI) technology has provided a new practical path for classroom activity design. Taking college English teaching in China as the research scenario, this paper focuses on AI-driven classroom activity design. By analyzing its core characteristics and teaching adaptability, it explores its effectiveness in enhancing learning participation, realizing personalized teaching, and optimizing the evaluation system, while examining practical challenges such as insufficient technology adaptation, homogeneous activities, and inadequate teacher literacy. On this basis, it proposes optimization paths including a "technology-goal-discipline" three-dimensional adaptation framework, a teacher literacy improvement system, and data ethics norms. The study holds that AI-based classroom activity design must take teaching goals as the core and balance technology application with the essence of teaching to provide sustainable support for improving the quality of college English teaching.

Research on the reconstruction and practice path of big data course teaching mode driven by generative artificial intelligence technology

Hao Sun — Thu, 02 Apr 2026 16:49:46 +0800

This paper explores the pedagogical transformation of big data courses through generative AI, first examining traditional teaching challenges: content-industry mismatch, inadequate alignment between practice and real-world scenarios, and limitations in instructional implementation and evaluation systems. It then proposes a restructuring framework: clarifying the division of responsibilities between human and machine collaboration, developing a dynamically evolving curriculum system based on knowledge progression, and implementing precision teaching with multidimensional assessment through student analytics. Finally, case studies from Beijing No.11 School and Khan Academy demonstrate generative AI applications in data tracking, knowledge delivery, and personalized instruction, offering actionable insights for advancing big data course reform.

Research on AI-driven social media recommendation strategies for traditional cultural content

Hui Wang, Haiying Liang — Thu, 02 Apr 2026 16:49:56 +0800

The algorithmic distribution of social media has changed the traditional pathways of cultural dissemination, with both trending and fragmented content coexisting, which can easily lead to symbolic consumption and shifts in value. This article takes a humanities-oriented approach, exploring methods such as semantic modeling, dynamic profiling, weight adjustment, collaborative production, and multimodal experiences to enhance the interpretability and transmission effectiveness of recommendations. It aims to provide strategic reference for platform governance and high-quality content supply, while promoting sustained cultural understanding among audiences during their routine browsing.

Tool path planning for NC machining of complex parts based on UG

Liu Yao — Thu, 02 Apr 2026 16:49:59 +0800

With the ever-increasing demands for machining accuracy and efficiency of complex parts in the high-end manufacturing industry, tool path planning has become a key link in numerical control (NC) machining, which directly affects the machining quality and production cost of parts. To improve the machining efficiency and surface quality of parts such as complex curved surfaces and thin-walled structures, this paper discusses the application of UG software in tool path planning. First, by utilizing the parametric modeling and machining modules of UG, a multi-strategy collaborative path generation scheme including contour layered machining, adaptive region segmentation and spiral cutting was designed for complex geometric features. Second, the genetic algorithm was integrated to optimize the tool path, so as to reduce non-cutting travel and abrupt steering. Finally, the feasibility and safety of the planned path were dynamically verified through the cutting simulation and collision detection functions of UG. By means of UG and knowledge fusion tools, intelligent algorithms are deeply integrated with the Computer-Aided Manufacturing (CAM) process, realizing the collaborative and autonomous optimization of tool paths and cutting parameters.

Dual PASNet predicts relevant markers and key pathways of LSCC

Xudan Zhou, Qinglin Yang, Yuxin Zhang, Xiaoli Chen, Jin Luo, Guohui Ma, Wei Shu — Thu, 02 Apr 2026 16:50:04 +0800

To identify core prognostic markers and key pathways of lung squamous cell carcinoma (LSCC), this study adopted the interpretable Dual PASNet deep learning model with a dynamic pathway mask mechanism to screen survival-related pathways/genes, and systematically validated their prognostic value through multi-cohort verification, feature selection, and functional analysis. The model identified 10 survival-related core pathways in LSCC (P<0.01), with p53 signaling, cell cycle, and PI3K-Akt signaling pathways having the highest weights. Sixteen core genes showed significant expression differences between high- and low-risk groups in both TCGA-LUSC (internal) and GSE19804 (external) cohorts (*P<0.001), and 10 prognostic signature genes were further screened by LASSO regression. The 10-gene prognostic model exhibited robust risk stratification (Log-rank P=0.000765), with time-dependent ROC AUC of 0.709–0.780 at 6–60 months. Functional enrichment indicated significant enrichment in tumor malignant phenotype-related pathways (cell cycle, DNA replication). Immune analysis showed a close association with LSCC immunosuppressive microenvironment (more prominent in high-risk group), which was validated by single-cell transcriptome analysis (GSE131907) showing specific expression of signature genes in myeloid cells and T lymphocytes. This 10-gene prognostic signature has reliable prognostic predictive value for LSCC, reveals key molecular regulatory pathways and immune microenvironment characteristics, and provides potential molecular markers and a theoretical basis for precise prognostic evaluation and targeted therapy of LSCC.

Artificial intelligence in higher education: A human-centered implementation framework

Yuan Kang, M. KazemChamran — Thu, 02 Apr 2026 16:50:07 +0800

Artificial intelligence (AI) is increasingly integrated into higher education through machinelearning (ML) and natural language processing (NLP) systems that support teaching, learning, student services, and institutional decision-making. Learning analytics models can enable earlier and more targeted student support, while adaptive and tutoring systems personalize practice and feedback. More recently, generative AI and large language models (LLMs) have introduced dialog-based assistance for explanation, feedback drafting, and study support at scale. However, these benefits come with nontrivial risks, including unreliable output, bias, privacy leakage, and uncertainty about acceptable use in assessment. This article presents a concise, human-centered implementation framework that links AI capability selection to learning outcomes, data governance, assessment design, and post-deployment monitoring, enabling institutions to deploy AI in ways that are pedagogically meaningful, accountable, and sustainable.

Research progress on strategies and mechanisms of interface engineering to enhance the stability of organic solar cells

Yangyang Yu — Thu, 02 Apr 2026 16:50:10 +0800

Organic solar cells (OSCs) have broad application prospects in fields such as photovoltaic building integration and wearable electronics due to their advantages of flexibility, lightweight, solution processability, and low cost. In recent years, their photoelectric conversion efficiency has exceeded 20%, laying the foundation for commercial development. However, insufficient stability remains the core bottleneck that restricts its practical application. Environmental factors such as light, heat, and humidity can easily lead to rapid efficiency degradation of the device, with interface defects causing charge recombination, interlayer delamination, and water oxygen permeation being the main causes. As a key hub for charge separation, transmission, and collection in OSCs, the performance of the interface directly determines the stability and long-term service capability of the device. Therefore, interface engineering has become a key path to break through the bottleneck of OSCs stability. This article systematically reviews the mainstream strategies for improving the stability of OSCs through interface engineering through literature research, analyzes the mechanisms of each strategy, summarizes current research progress and existing problems, and proposes future prospects that fit the perspective of undergraduate research. It provides a basic reference for the preparation of high stability OSCs and promotes their commercialization process.

Remote sensing image classification based on deep learning

Peng Yin — Thu, 02 Apr 2026 16:50:12 +0800

Remote sensing image classification is a core component of remote sensing technology applications, widely serving fields such as resource exploration and environmental monitoring. Traditional classification design schemes have limitations such as insufficient feature extraction and low classification accuracy, making it difficult to adapt to complex scene requirements. Deep learning, with its powerful ability to mine deep features, provides a new design concept for remote sensing image classification. This article focuses on the design of a remote sensing image classification system based on deep learning, reviews the relevant technical foundations, completes the design of each core module and overall architecture, solves the pain points of traditional design, and provides efficient and feasible design solutions for remote sensing image classification.

Intelligent diagnosis system for rice diseases and pests based on keras model and IoT

Yan Li, Genjuan Ma, Yuwei Gu, Yuqing Zhang, Ziqing Yang, Jiahao Cao, Jiawen Yin — Thu, 02 Apr 2026 16:50:19 +0800

As a major rice-producing country, China's rice yield and quality are critical to food security and livelihood protection. Rice diseases and pests have become a key constraint on rice production, reducing both yield and quality and seriously affecting farmers' income and food supply stability. To realize accurate prevention, real-time monitoring and scientific management of rice diseases and pests, and overcome the low efficiency, poor accuracy and complex operation of traditional monitoring methods, this paper designs an intelligent diagnosis system for rice diseases and pests based on the Keras model and Internet of Things (IoT). The system integrates lightweight edge computing devices, UAV inspection terminals and environmental sensors, taking the lightweight Keras deep learning model as the core to achieve rapid identification and accurate diagnosis. With IoT and edge intelligence, it realizes real-time collection, monitoring and analysis of field environmental parameters such as temperature, humidity and light intensity, supporting intelligent monitoring, data recording and dynamic management throughout the rice growth cycle. The system forms a complete digital and intelligent monitoring and diagnosis scheme with standardized processes, providing efficient and scientific management support for farmers and effectively improving the prevention and control efficiency of rice diseases and pests.

Semantic-aware network collaborative scheduling mechanism for multi-agents

Yining Gao — Thu, 02 Apr 2026 16:50:22 +0800

In the context of the deep integration of intelligent terminals and Internet of Things technology, the multi-agent semantic perception network, formed by the combination of multi-agent systems and semantic perception technology, is widely applied in various scenarios such as distributed information processing and intelligent collaborative perception. The efficiency of its collaborative scheduling plays a decisive role in the network's perception performance and service quality. This paper summarizes the basic characteristics of the multi-agent semantic perception network, analyzes the core problems existing in the current collaborative scheduling process, proposes an appropriate collaborative scheduling optimization mechanism, and utilizes semantic unified adaptation and dynamic scheduling strategies to resolve issues such as semantic heterogeneity and unreasonable resource allocation, thereby improving the network's collaborative efficiency and stability. This enables the network to operate more efficiently and stably in practical applications, providing strong support for the development of related fields.

Analysis and simulation verification of key factors affecting safe rescue in building fire scenarios

Xinzhe Zhao, Yunqi Jing, Huixin Dai, Yuming Kuang, Chongtian Wan — Thu, 02 Apr 2026 16:50:25 +0800

To address the issues of low efficiency, prominent safety risks, significant congestion interference, and communication disruption in building fire rescue operations, this paper proposes a rescue optimization model that integrates multiple constraints. The model takes "efficiency-safety-balance" as its core objective, incorporates the Two-in-Two-out safety rule, congestion coefficient, and the Bernoulli distribution of communication failure, quantifying the trade-offs among multiple objectives. Through Monte Carlo simulation and multi-scenario parameter analysis, the robustness of the model in 3-15 story building scenarios is verified, revealing the coupling influence mechanism of communication failure rate and risk threshold on rescue performance. The results show that the model can significantly reduce the total rescue time and safety violation rate, providing quantitative decision support and practical solutions for emergency dispatch in high-rise buildings.

Practical research on aerobic exercise promoting the physical fitness of college students

Long Zhao — Thu, 02 Apr 2026 16:50:28 +0800

Aiming at the problems of declining physical fitness and insufficient participation in sports activities among contemporary college students, this study takes four types of aerobic exercises (rope skipping, jogging, yoga, and swimming) as the carriers, combines the innovation of intelligent equipment with the optimization of teaching models, and explores the practical effect of aerobic exercise on improving college students' physical fitness. Adopting literature research method, field investigation method, group experiment method and other research approaches, the study conducted physical fitness assessment and demand analysis on 300 college students, and designed a practical model of "Personalized Program + Intelligent Monitoring + Interactive Incentive". Through 1 to 3 months of intervention experiments, the significant effectiveness of innovative aerobic exercise in improving core physical fitness indicators such as cardiopulmonary function, endurance and flexibility was verified. The results show that scientific aerobic exercise intervention can increase college students' vital capacity by 5%~15% and endurance indicators by 5%~12%, with the exercise adherence rate reaching more than 70%. The integration of intelligent equipment and interesting exercise forms can effectively stimulate students' interest in participation and solve core problems such as irregular exercise and unscientific training methods. The closed-loop model of "Physical Fitness Assessment - Program Customization - Dynamic Adjustment - Achievement Transformation" constructed in this study can provide a reference for the reform of college physical education teaching and the implementation of the National Fitness Strategy in colleges and universities.

Research on user intent perception and experience optimization in multimodal intelligent interaction scenarios

Ziyou Ma — Thu, 02 Apr 2026 16:50:31 +0800

To address the issues of low accuracy in intent recognition and poor user experience in singlemodal interaction in complex scenarios, this paper conducts research on user intent perception and experience optimization in multi-modal intelligent interaction scenarios. The research integrates voice, gesture, facial recognition, and environmental perception technologies, builds a feature fusion model based on Convolutional Neural Network (CNN) Transformer, designs a hierarchical intent understanding mechanism and interaction execution optimization strategy, and realizes precise perception and efficient response of user intentions. Experiments are selected to verify in three typical scenarios: smart home control, immersive entertainment, and personalized services. The results show that this research scheme can achieve an intent recognition accuracy of over 85%, and the system response delay is controlled within 800ms. This effectively improves the naturalness and intelligence level of multi-modal interaction, providing technical references for the design and optimization of intelligent terminal interaction systems.

Intelligent patrol and targeted therapy integrated breast cancer defense nanorobot: Technology, application, and prospects

Haidi Yu — Thu, 02 Apr 2026 16:50:35 +0800

Breast cancer, with the highest morbidity and mortality among women globally, presents significant challenges in early diagnosis and precision treatment. This paper proposes a breast cancer defense nanorobot integrating intelligent patrol and targeted therapy functions. The nanorobot incorporates biosensors, targeting technology, drug delivery systems, and intelligent navigation control, enabling highly sensitive identification, precise localization, and controlled drug release. Structurally, it utilizes DNA origami and biodegradable materials for biocompatibility and stability. Functionally, it combines FRET sensing with CRISPR gene editing to enhance diagnostic accuracy and therapeutic efficacy. Through multidisciplinary collaboration, the system underwent full-chain development from design to experimental validation. Results demonstrate excellent motility, targeting ability, and therapeutic effect, showing broad application prospects in breast cancer treatment.

Drug screening based on molecular fingerprint similarity

Luo Hong — Thu, 02 Apr 2026 16:50:37 +0800

The structural characterization integrity of molecular fingerprints directly impacts the efficiency and accuracy of drug molecule screening. Single molecular fingerprints suffer from limitations such as incomplete structural representation and insufficient generalization capabilities. To address this challenge, this study employs three complementary molecular fingerprints (MACCS, PubChem, Pharmacophore ErG) to construct a multidimensional molecular characterization system. Molecular similarity is calculated to screen the top 5 candidate molecules by similarity ranking. Subsequently, the screened candidate molecules undergo molecular docking validation with the 4mbs receptor to assess binding affinity and drug-like potential. Experimental results demonstrate that the five selected candidate molecules exhibit high binding affinity and excellent docking scores with the 4mbs receptor, while maintaining good consistency with the query molecules in both chemical structure and molecular fingerprint similarity. This study indicates that the proposed method effectively overcomes the limitations of single fingerprint approaches, significantly enhancing the accuracy and robustness of drug molecule screening. This methodology provides efficient and reliable technical support for candidate molecule selection in the drug discovery process.

Research on the optimization of interactive strategies in intelligent classrooms of higher vocational business English empowered by AI+BOPPPS

Xiaolan Zhu, Jing Feng — Thu, 02 Apr 2026 16:50:41 +0800

This paper focuses on classroom interaction in vocational business English teaching under educational digital transformation. Traditional classrooms are limited by monotonous interaction, low participation, weak personalization, and delayed feedback, which can hardly meet the needs of cultivating interdisciplinary talents. Supported by TAM and constructivist theory, this study integrates AI and the BOPPPS model to construct an optimized teaching strategy, verifies its effect through empirical research, and puts forward supporting measures. It provides theoretical and practical references for the digital, intelligent, and personalized reform of vocational business English education.

Security risk analysis and countermeasure technologies for unmanned aerial vehicles

Yishi Xue — Thu, 02 Apr 2026 16:50:45 +0800

The expanding deployment of unmanned aerial vehicles (UAVs) across logistics, agriculture, infrastructure inspection, and emergency response has generated corresponding growth in security incidents. This paper presents a systematic survey of UAV security threats and their associated countermeasure technologies. A three-layer risk taxonomy is proposed that classifies threats into physical layer risks, communication layer risks, and data and privacy risks. The cross-layer cascade propagation mechanisms through which localized anomalies escalate into systemic failures are also analyzed. For each risk category, state-of-the-art countermeasures are reviewed, including multi-modal sensor fusion detection, graduated-response counter-UAV systems, network security hardening, digital-twin-enabled airspace governance, and AI-augmented intelligent management. The analysis reveals that effective UAV security governance requires a cross-layer defense-in-depth architecture coordinating detection, prevention, response, and regulatory enforcement.

Research on intelligent monitoring system for safety use of large steel structure building

Wenhao Lu, Zizhen Shen — Thu, 02 Apr 2026 16:50:47 +0800

To address the whole-lifecycle safety monitoring needs of large steel structures, this study developed an intelligent monitoring system integrating a four-layer architecture of perception, transmission, platform, and application. By optimizing the deployment of sensor networks, it achieved synchronous multi-source data acquisition of critical parameters including stress and deformation during construction, dynamic response during operation, and fatigue damage during the aging phase. A data-driven model was developed for structural state identification and safety assessment, establishing multi-level warning mechanisms and methods for predicting remaining service life. Engineering applications demonstrate that this system effectively enhances structural safety monitoring accuracy and warning timeliness, providing technical support for the whole-lifecycle safety management of steel structures.

Interactive strategy optimization in higher vocational business English blended teaching empowered by AI and BOPPPS in the digital intelligence era

Jing Feng, Xiaolan Zhu — Thu, 02 Apr 2026 16:50:52 +0800

In the era of digital intelligence, artificial intelligence (AI) provides strong technical support for the innovation of business English teaching in higher vocational colleges. Aiming at the problems of insufficient interaction, single evaluation mode, and weak practical application ability in traditional business English teaching, this paper constructs an interactive strategy system of "AI BOPPPS" blended teaching. Based on the six links of the BOPPPS model, namely Bridge-in, Objective, Pre-assessment, Participatory Learning, Post-assessment, and Summary, this study integrates AI technologies such as intelligent teaching platforms, AI oral evaluation, and big data analysis to realize multi-dimensional interactions including teacher-student interaction, student-student interaction, and human-computer interaction. Empirical results show that the strategy significantly improves students' learning participation, autonomous learning ability and business English application level. This research provides a referable teaching mode and practical path for the blended teaching reform of higher vocational business English.

AI-based dynamic optimization of smart traffic signals for improved mobility efficiency

Jiangeng Long, M. KazemChamran — Thu, 02 Apr 2026 16:50:54 +0800

Artificial intelligence (AI) can improve urban mobility by adapting traffic signals to changing demand, incidents, and multimodal priorities. We propose a fieldable framework combining real-time state estimation, short-horizon prediction, and constraint-aware reinforcement learning. The policy suggests phase, split, and offset updates, while a safety and fairness layer enforces minimum green, clearance, pedestrian service, and side-street protection. Neighborhood messaging enables corridor coordination, and deployment uses monitoring and rollback. Illustrative studies show reduced delay and queues with higher throughput.