最新刊期
卷 71 ， 期 4 ， 2025
- “智能技术发展推动数字化学习，认知负荷智能化评估成研究热点。专家梳理机器学习、深度学习算法在评估中的应用，探索大语言模型潜力，为高效学习提供帮助。”摘要：With the rapid development of intelligent technology, the gradual popularization of digital learning has provided learners with diversed resources and pathways to learn. However, while this learning mode enriches the learning experience, it also brings complex challenges to learners' cognitive load. Traditional cognitive load measurement methods have problems with accuracy, such as insufficient processing, a lack of dynamic monitoring, and a lack of immediate feedback. Therefore, intelligent assessment of cognitive load has gradually become a popular research topic. Intelligent assessment can achieve accurate assessment of cognitive load by combining physiological indicators and using non-invasive technical means such as machine learning and deep learning. This study systematically reviews the applications of traditional machine learning algorithms, such as Support Vector Machine (SVM), Random Forest, and Linear Discriminant Analysis (LDA), as well as deep learning algorithms, including Feedforward Neural Networks (FNN), Convolutional Neural Networks (CNN), and Long Short-Term Memory networks (LSTM), in the intelligent assessment of cognitive load. It also explores the applications of the hybrid models. Finally, the study proposes insights and development directions for the intelligent assessment of cognitive load in digital learning and explores the potential of the large language model(LLM) in the intelligent assessment of cognitive load, aiming to provide assistance for efficient digital learning.关键词：digital learning;cognitive load;intelligent assessment;machine learning;deep learning100|64|0更新时间：2025-08-28
- “在室内定位领域，可见光惯性紧组合定位系统有效提升了遮挡和倾斜情况下的定位精度，平均精度达10.27厘米。”摘要：Visible light positioning (VLP) has received widespread attention in indoor positioning research and applications due to its advantages such as low cost, high accuracy, and extensive deployment in existing infrastructure. However, VLP systems experience a significant decline in localization performance when obstructed or tilted at the receiver end. To address this issue, a VLP/INS tightly coupled navigation system enhanced by adding motion constraints is proposed, which uses an inertial measurement unit for inclination estimation and applies filtering and fusion techniques at the raw observation level of the sensor to mitigate the effects of obstructions. In practical tests, the system achieved an average positioning accuracy of 10.27 cm, with accuracy improvements of 35.34% and 63.19% compared to pure VLP under occlusion and tilt situations, respectively, resulting in higher accuracy and robustness than existing methods.关键词：indoor positioning;visible light positioning;inertial navigation;tightly-coupled integrated system;motion constraints54|60|0更新时间：2025-08-28
- “在交通标志检测领域，研究者提出了基于改进YOLOv8模型的新方法，通过引入注意尺度序列融合机制和增加小目标检测层，有效提升了多尺度信息提取能力和小目标检测性能。同时，采用RT-DETR检测头和inner-mpdiou损失函数，显著提高了模型训练效率和精度。实验结果表明，该方法在保证实时性的前提下，平均精度高达84.0%，相较于YOLOv8模型提高了7.1%，整体模型大小降低了12.9%，有效提升了低分辨小目标检测的有效性。”摘要：Current methods for traffic sign detection primarily rely on single-stage deep learning algorithms to construct target detection models, which suffer from low detection accuracy and weak model generalizability. To address these issues, a traffic-sign detection method based on an improved the YOLOv8 model was proposed. This method introduces an attention-scale sequence fusion mechanism that enhances the ability of the neural network to extract multi-scale information. The addition of a small-object detection layer makes this method more suitable for small-object detection. Additionally, it adopts the RT-DETR detector head to process multi-scale features efficiently by decoupling the intra-scale interactions and cross-scale fusions. Moreover, to overcome the limitations of existing methods in terms of weak generalization and to improve the accuracy and efficiency of bounding box regression, a novel loss function， inner-mpdiou， is employed, which effectively improves the training efficiency and accuracy of the model. Experimental results on the Tsinghua-Tencent 100K (TT100K) dataset showed that under the premise of real-time performance, this method achieved an average accuracy of up to 84.0%. Compared with the current international mainstream YOLOv8 model, its average accuracy was improved by 7.1%, and the overall model size was reduced by 12.9%, thereby enhancing the effectiveness of low-resolution small object detection.关键词：small-object detection;traffic-sign detection;YOLOv8;low resolution;loss function317|223|0更新时间：2025-08-28
- “GMambaScanX在三维人体姿态估计领域取得突破，通过GCN和Mamba双流并行方法，有效提取人体结构特征，提高模型效率。”摘要：Transformer-based models have excelled in 3D human pose estimation tasks by effectively encoding the interconnections among all joints, thus capturing the global dependencies inherent in the data. However, such methods often fail to adequately model the local dependencies of joints and suffer from a quadratic increase in computational complexity with sequence length, leading to a waste of computational resources. To address these issues, this study defines the human motion process as a sequence of input-output processes of a state-space model and proposes a 3D human pose estimation method, GMambaScanX, based on the fusion of GCN and Mamba. The GCN model captures the spatiotemporal local dependencies of human motion sequences, thereby enhancing the model's ability to capture local relationships. The Mamba model captures the spatiotemporal long-range dependencies of human motion sequences, thereby improving the model's ability to capture global relationships. We present two MambaScan scanning strategies: a sequential joint scan for temporal motion state transitions, which enhances the model's comprehension of human motion features within the temporal dimension, and a spatial joint scan informed by human structural prior knowledge, which augments the model's understanding of human structural features in the spatial domain. GMambaScanX was trained and validated on the Human3.6M dataset, achieving an average joint position error of 39.8 mm with only 12.4% (5.3×106) of the parameters of MotionBERT. Compared with fully connected models, GMambaScanX better extracts human structural features, utilizes parameters more efficiently, and improves the model's effectiveness and efficiency.关键词：3D human pose estimation;state-space model;graph convolutional network;human joint structure scanning112|93|0更新时间：2025-08-28
- “在人体姿态估计领域，MMSF模型通过多尺度与多级语义融合，提高了估计精度，降低了模型复杂度。”摘要：To addressthe issues of limited scale diversity and neglect of close-range information in visual Transformer models for human pose estimation tasks, this sutdy proposes a multiscale and multi‑level semantic fusion transformer (MMSF) model. This model realizes the mutual learning of visual information at different resolutions through the introduction of keypoint tokens as proxy cross-Transformer operations, thereby improving the estimation accuracy. Meanwhile, by utilizing deep convolution and DenseNet’s token reuse technique, it effectively extracts cross tokens containing multi‑level semantic information, reducing the stacking of encoder layers and model complexity. The integration of multi‑level semantic information is achevied through the cross-fusion attention operation involing cross tokens and standard tokens, thereby further enhancing the performance of pose estimation. The experimental results show that, under the same conditions, the MMSF model achieves an average precision of 78.1% on the COCO dataset, which is 2.3% higher than the TokenPose baseline model, and has also demonstrated its effectiveness on the MPII dataset, outperforming classical Transformer-based human pose estimation methods in recent years.关键词：vision Transformer;human pose estimation;deep convolution;token fusion;cross attention197|149|0更新时间：2025-08-28
- “在不实信息核查领域，研究者提出了融合大语言模型和证据抽取的事实核查模型，有效提升了事实核查的准确度。”摘要：Fact-based verification is currently the mainstream method used in research on misinformation verification. However, existing research results have issues such as the low relevance between the document content extracted during document retrieval and the statements to be verified, the neglect of the intrinsic connections between sentences in evidence retrieval, and the insufficient logical reasoning ability of small language models in statement verification. To address these issues, a fact-checking model that integrates large language models and evidence extraction has been proposed. To enhance the relevance between the statements to be verified and the document content, a document retrieval for claim verification（DRCV） document retrieval algorithm was introduced. To extract the most relevant sentences from documents as evidence, a “document-statement pair” training evidence retrieval model was constructed, along with a keyword-attention mechanism-based evidence retrieval method. To improve the model’s logical reasoning capability and increase the accuracy of fact-checking, a statement verification model based on large language models was developed. This model selects seven large language models with parameter sizes ranging from 500 million to 130 billion to verify statements and utilizes their logical reasoning abilities to check the facts of statements. Simulation experiments was conducted on real datasets to evaluate the proposed fact-checking model. The results indicate that the accuracy of fact-checking using this model is 0.1% to 34.0% higher than that of models using only large language models, and it surpasses the accuracy of the currently best-performing model by 1.8%.关键词：misinformation detection;evidence extraction;fact verification;large language model196|241|0更新时间：2025-08-28
- “在社交媒体大数据背景下，专家提出了基于语义引导注意力和多任务学习的图文情感分析方法，为理解用户情感倾向提供解决方案。”摘要：In the backdrop of today's big data milieu on social media and online platforms, image-text sentiment analysis has become an important research task, which is crucial for understanding users' emotional tendencies. Existing methods are usually limited to the single-level features of modalities, lack an understanding of multi-level emotional information in images, and are prone to information redundancy and feature shift during multimodal feature fusion, resulting in poor model performance. In response to these issues, this study proposes an image-text sentiment analysis based on semantic guided attention and multi-task learning. Capture multilevel emotional information of images through a multiscale feature extraction module, use semantic-guided attention to fuse image information related to textual emotional information, and introduce an emotional focus calibration task in the multi-task learning module to minimize the distance between the fused features and their emotional centroids. Experimental results obtained from three social media datasets demonstrate that the proposed method outperforms existing methods in image and text sentiment analysis tasks.关键词：image-text sentiment analysis;multi-scale feature extraction;semantic guided attention;motional focus calibration;multi-task learning169|246|0更新时间：2025-08-28
- “在文本情感分类领域，专家提出了基于BERT的多情感字符级信息注意力方法CBSA，通过整合字符级信息和情感来源信息，显著提高了情感预测准确性。”摘要：Current deep learning-based methods for text sentiment classification have not fully exploited character-level information and sentiment source information, leaving room for improvement in the field. To address this, we propose a novel method, Character-BERT-Sentiment-Attention (CBSA), which leverages a pretrained BERT model and focuses on multi-sentiment character-level information. This approach enhances the semantic representation of text by capturing character-level features and concatenating BERT’s pretrained word embeddings of BERT to form a unified semantic vector. Subsequently, contextual words are integrated with four types of sentiment source information (sentiment words, negation words, degree adverbs, and conjunctions) into a gated recurrent unit（GRU) neural network. An attention mechanism is employed to generate a sentence representation that incorporates sentiment-related resources, followed by a softmax classification layer to predict sentiment polarity. By extracting both character-level and multi-sentiment source information, the proposed method captures sentiment-relevant features from diverse representational subspaces, thereby improving the sentiment prediction accuracy. Experimental results on five benchmark datasets demonstrate that the proposed method significantly improves classification accuracy.关键词：sentiment classification;Bert;character-level information;fusion vector;attention139|11|0更新时间：2025-08-28
- “在加密DNS通讯领域，EDGAD方案提升了DGA流量检测的实时性、准确率和泛化性，有效识别7种DGA恶意软件，准确率达到98.07%。”摘要：An encrypted traffic detection scheme (EDGAD) is proposed to enhance the real-time detection, accuracy, and generalization of domain generation algorithm (DGA) traffic in encrypted domain name system (DNS) communications and ensure computer communication security. EDGAD comprises two modules: data preprocessing and traffic classification. The preprocessing module employs traffic clustering and an efficient feature set to reduce the overhead. The traffic classification module includes a two-stage model. In the first stage, a binary classifier is used to distinguish DNS from HTTPS (DoH) traffic as either DGA or non-DGA. The second stage employs contrastive learning to construct a multiclass model that identifies the specific DGA software that generates the encrypted traffic. This stage also incorporates traffic sample enhancement and the joint optimization of contrastive learning and multi-classification modules to improve generalization and training efficiency. Experiments using 1 s traffic data indicated that the XGBoost model was optimal for the first-stage binary classification, with the second-stage scaling parameter set to 0.25. The results demonstrate that EDGAD effectively identifies seven types of DGA malware, with an accuracy of 98.07%, and an average precision mean value of 0.981 3, and an improvement of 1.24 percentage points and 0.013 6 compared with the comparison scheme, respectively.关键词：machine learning;traffic analysis;anomaly detection137|227|0更新时间：2024-11-11
- “在量子图像加密领域，专家提出了基于等比数列的高维广义Arnold变换量子图像置乱算法，验证了其可行性，具有较大密钥空间和抗攻击能力，满足密码学要求。”摘要：To address the limitations of inflexible coefficients in low-dimensional regular matrix encryption algorithms and the challenges associated with constructing high-dimensional encryption matrices, this study proposes a high-dimensional generalized Arnold transform for the quantum image scrambling algorithm based on integer matrices obtained from geometric sequences. Initially, two high-dimensional integer matrices with unit determinants were constructed, and a high-dimensional generalized Arnold transform matrix was derived using conventional matrix multiplication techniques. Subsequently, by employing a universal color quantum image representation framework, this transformation matrix was seamlessly integrated into the quantum image encryption process. Additionally, the inverse of the high-dimensional generalized Arnold transformation matrix is formulated and utilized in the image decryption procedure based on a universal-color quantum image representation. The proposed algorithm boasts a diverse range of transformation formulas that enable the generation of high-dimensional encryption matrices. The feasibility of this approach is exemplified by the encryption of 24-bit true-color images. The simulation results underscore the algorithm’s expansive key space, enhanced key randomness, and robust anti-attack capabilities, thereby fulfilling the stringent requirements of cryptography and demonstrating significant theoretical and practical merits.关键词：quantum image processing;quantum image encryption;quantum image representation model;high-dimensional geometric transformation;geometric sequence;generalized Arnold transformation45|42|0更新时间：2025-08-28
- “最新研究进展表明，基于邻域信息与嵌套注意力的知识图谱补全方法，有效提升了工业领域模型预测准确性，同时保持较低参数量。”摘要：With the increase in the complexity of industrial systems and the demand for intelligence, knowledge graphs are being increasingly applied in industrial fields. Knowledge graph completion is essential for optimizing processes, enhancing decision quality, and enabling automation. However, most current methods focus only on node similarity and overlook the rich information in node neighborhoods, thereby limiting the relationship inference. To address this issue, we propose a knowledge graph completion method based on neighborhood information and nested attention. Our proposed model calculates the neighborhood information for all triples, embeds it to form neighborhood and triple representations, and encodes them using a nested attention mechanism to maintain the stability of the structural features. Finally, an inner product decoder completes the missing relationships. The experiments were conducted using three datasets, and the results showed that the prediction accuracy of the model improved while maintaining a low number of parameters. This provides a method for knowledge graph completion tasks that balances model scale and prediction accuracy.关键词：knowledge graph;knowledge graph completion;knowledge reasoning;Transformer;attention mechanism39|24|0更新时间：2025-08-28
- “在芯片点胶检测领域，专家改进YOLOv8n模型，提出SPA-KG算法，提升检测速度和精度，满足工业需求。”摘要：To address the problems of slow speed, low accuracy, and insufficient inference ability in the detection of bad appearance under chip dispensing motion, the original YOLOv8n model was improved. Based on the improved YOLOv8n, a detection algorithm called Self-Position Attention-Knowledge Graph (SPA-KG), which fuses knowledge graphs and improves YOLOv8n, was proposed. First, SPA attention is designed based on coordinate attention (CA), and SPA attention can learn detailed information about small targets more fully. Second, lightweight convolution modules GHOSTConv and Adaptive Kernel Convolution (AKConv) were introduced into the backbone network and feature fusion networks, respectively, and spatial pyramid pooling was improved by using Simplified Spatial Pyramid Pooling-Fast (SIMSPPF). The number of parameters in the model was reduced to improve the detection speed, and the α-EIoU loss function was design by combining the α-IoU and EIoU to improve the localization ability and recognition accuracy of the algorithm. The experimental results showed that the average precision of the SPA-KG reached 96.7%, the precision rate reached 94.2%, and the recall rate reached 94.0%. The number of parameters reached 2.58×106, and the detection speed reached 107.7 frames/s. SPA-KG meets the industrial detection requirements.关键词：appearance inspection;knowledge graph;YOLOv8n;attention mechanism;machine vision43|39|0更新时间：2025-08-28
- “在水资源保护领域，SSP-GLM模型通过深度学习技术，有效提升了再生水排放指标预测的准确性，为工业污水处理智能化管理提供技术支持。”摘要：As a vital facility for water resource protection, monitoring and predicting the discharge indicators of reclaimed water is one of the core tasks of a wastewater treatment plant. Given the high complexity and poor scene consistency of wastewater discharge data, traditional time-series prediction methods are limited in their effectiveness. To address this, a novel industrial time-series prediction model, the Sewage-oriented Serial data Processing Generative Large Model (SSP-GLM), was proposed. This model segments time-series data into subsequences, extracts local features using deep convolutional neural networks (CNNs), and performs time-series reasoning with a generative large model (GLM), thereby enhancing the prediction accuracy. Experiments were conducted using real effluent data from two wastewater treatment plants in Xi’an, with the mean squared error (MSE) and mean absolute error (MAE) as evaluation metrics. The results show that SSP-GLM outperforms baseline methods, such as GRU, DLinear, and Autoformer, in both full-sample and few-shot learning scenarios. In particular, under few-shot conditions, SSP-GLM demonstrates significantly stronger capabilities in capturing complex temporal features. The model also exhibited good generalization across different wastewater treatment plants, providing robust technical support for the intelligent management of industrial wastewater treatment.关键词：industrial software;wastewater treatment indicator prediction;generative large language model;time-series pattern prediction105|43|0更新时间：2025-08-28
- “在水力发电领域，专家提出了基于大语言模型与马尔可夫概率逻辑生成式神经网络的智能分析方法，为水电站大坝结构设计和材料选取提供参考。”摘要：Aiming at the international frontier issue of structural deformation in hydropower facilities located in high mountains and valleys under extremely complex conditions during long-term operation, and based on the analysis of the causes of dam deformation characteristics and the study of the long-term operational characteristics of hydropower station dams, a generative intelligent analysis method combining large language models (LLMs) with Markov probabilistic logic generative neural networks has been proposed. This method leverages the powerful semantic understanding and generative capabilities of LLMs to optimize the feature extraction and correlation analysis of hydropower station dam deformation data, thereby enhancing the model accuracy in addressing the complex interactions of multiple factors under extreme conditions. This method was applied to the main project of the Dadu River Pubugou Hydropower Station dam infrastructure, which is affected by extreme weather, geological disasters, and sudden changes in turbulent flow of the Dadu River. It is used for the precise prediction and early warning of potential major hazards, such as the displacement of the dam’s main structure, special skeletons, and large-scale equipment. Extensive experimental verification has shown that under extremely complex conditions, the Markov probabilistic logic generative neural network combined with large language models and knowledge graphs can accurately identify the non-uniform deformation characteristics of hydropower station dam infrastructure and uncover the interrelationships between the deformation processes of various dam parts due to internal force interactions. Compared with the knowledge graph embedding methods, the overall accuracy of the results was significantly improved. This breakthrough provides a reference for the structural design and material selection of various parts for constructing of future hydropower stations in China.关键词：large language model;knowledge graph;Markov probability;hydropower station infrastructure;extremely complex conditions;generative intelligence;sensor network;knowledge graph embedding41|11|0更新时间：2025-08-28
- “在时间序列预测领域，研究者提出了mWKN-HAGRU时序预测方法，显著提升了船舶柴油机冷却水系统性能参数预测性能。”摘要：To improve the feature extraction and prediction ability of gate recurrent unit (GRU) for time series, we propose a performance parameter prediction method based on a multi-scale wavelet kernel network and a hybrid attention-gated recurrent network (mWKN-HAGRU). First, the scaling parameters of the wavelet kernel function ware adjusted to extract hidden state information from multiple dimensions, providing rich feature inputs for subsequent predictive models. Then, a hybrid attention gate recurrent unit (HAGRU) is designed, where temporal local attention can learn the long-term dependence between sequences and quantitatively characterize the influence of different features on the prediction performance.This can effectively capture the information interactions between features and comprehensively characterize the spatial correlation and evolution laws between different sequences. Finally, experimental results on a real marine diesel engine dataset showed that the proposed method significantly improved the predictive performance of the key performance parameters of the cooling water system.关键词：multi-scale wavalet kernel;hybrid attention;gated recurrent unit;time series prediction86|110|0更新时间：2025-08-28
- Tel：027-68756952 Email：whdz@whu.edu.cn
- Technical support is provided by Beijing Founder electronics co., LTD 京ICP备09064830号-19
京公网安备11010802024621 - It is recommended to read the content of this site in Chrome&IE9+. Please switch to extreme mode in browser 360.
- Cookies We use cookies to help provide and enhance our service and tailor content. By continuing, you agree to the use of cookies.
- Total Visits：1135748 Visits Today：161
Bulk Citation
0