检索结果-内蒙古大学图书馆

arXiv 2025年

作者： Wu, Zheng Xie, Yiping Zhao, Bo He, Jiguang Luo, Fei Deng, Ning Yu, Zitong School of Computing and Information Technology Great Bay University Dongguan523000 China Dongguan Key Laboratory for Intelligence and Information Technology Dongguan523000 China Computer Vision Institute School of Computer Science & Software Engineering Shenzhen University Shenzhen518060 China

Heart rate (HR) estimation via remote photoplethysmography (rPPG) offers a non-invasive solution for health monitoring. However, traditional single-modality approaches (RGB or Radio Frequency (RF)) face challenges in balancing robustness and accuracy due to lighting variations, motion artifacts, and skin tone bias. In this paper, we propose CardiacMamba, a multimodal RGB-RF fusion framework that leverages the complementary strengths of both modalities. It introduces the Temporal Difference Mamba Module (TDMM) to capture dynamic changes in RF signals using timing differences between frames, enhancing the extraction of local and global features. Additionally, CardiacMamba employs a Bidirectional SSM for cross-modal alignment and a Channel-wise Fast Fourier Transform (CFFT) to effectively capture and refine the frequency domain characteristics of RGB and RF signals, ultimately improving heart rate estimation accuracy and periodicity detection. Extensive experiments on the EquiPleth dataset demonstrate state-of-the-art performance, achieving marked improvements in accuracy and robustness. CardiacMamba significantly mitigates skin tone bias, reducing performance disparities across demographic groups, and maintains resilience under missing-modality scenarios. By addressing critical challenges in fairness, adaptability, and precision, the framework advances rPPG technology toward reliable real-world deployment in healthcare. The codes are available at: https://***/WuZheng42/CardiacMamba. Copyright © 2025, The Authors. All rights reserved.

关键词： Fast Fourier transforms

来源：评论

学校读者我要写书评

暂无评论

Infinity-Norm Weighted Low-Order Sparse Iterative DOA Estimation under Impulsive Noise

Infinity-Norm Weighted Low-Order Sparse Iterative DOA Estima...

引用

2023 IEEE International Conference on Signal Processing, Communications and Computing, ICSPCC 2023

作者： Zhang, Yongqing Liu, Zhiqiang Wang, Weidong Li, Hui Shi, Wentao School of Physics and Electronic Information Engineering Henan Polytechnic University Jiaozuo454003 China School of Computer Science and Technology Henan Polytechnic University Jiaozuo454003 China School of Software Henan Polytechnic University Jiaozuo454003 China Ocean Institute Northwestern Polytechnical University Taicang215400 China

ISBN: (纸本)9798350316728

This paper investigates the subject of the direction of arrival (DOA) estimation with acoustic vector sensor array (AVSA) under impulsive noise. A low-order-based variant of the typical sparse reconstruction method is designed to address this challenge. The proposed scheme successively weights the array output snapshots using the infinity norm to remove high outliers. Subsequently, the Frobenius norm based approach is extended to lower order, and a new cost function is created to address the impulse characteristics of the residual. Finally, the expression of the signal power is quantified, and DOA is approximated on the signal power by spectral peak search mechanism. Numerical results demonstrate that the proposed algorithm offers significant performance enhancements over some current techniques in impulsive noise scenarios. © 2023 IEEE.

关键词： Impulse noise

来源：评论

学校读者我要写书评

暂无评论

An Architectural Approach for Enhanced Data Interoperability Across Building Systems

An Architectural Approach for Enhanced Data Interoperability...

引用

Cloudification of the Internet of Things (CIoT)

作者： Peter Yefi Sikandar Ejaz Ramanunni Parakkal Menon Ursula Eicker Yann-Gaël Guéhéneuc Computer Science and Software Engineering Dept. Concordia University Montreal Canada Institute for Information Systems Engineering Concordia University Montreal Canada German Aerospace Center Institute for Solar Research Jülich Germany NextGen Cities Institute Concorida University Montreal Canada

ISBN: (数字)9798350356717

ISBN: (纸本)9798350356724

Access to building data is crucial for creating portable applications and improving building operation and energy efficiency. Ensuring data transmutability is essential for facilitating research and overcoming the diverse data representation, management, and collection methods across various building management systems (BMSs). Different BMSs use multiple interfaces for data exchange. Some BMSs provide Application Programming Interfaces (APIs) for data exchanges, while others have gateways connecting to cloud services for application subscription and data access. This study introduces a five-layered architecture that describes how client applications, researchers, and other interested stakeholders can exchange data with different BMSs. Our objective is to help researchers and practitioners understand the various ways of accessing transmutable (capable of being transformed into a compatible format) data from BMSs, regardless of the building management system a building uses. Through transmutable access to building systems’ data, we aim to enable transferable energy-efficient-related applications for buildings. We evaluate our work with a building model of the Varennes Library in Varennes, QC, Canada. The building model includes information about the building, floors, rooms, electric meter and historical time series data of the meter, weather station and its historical time series data. The model also has carbon dioxide concentration sensors, temperature sensors, humidity sensors, and historical time series data of the sensors. We include two test client programs to show how they access the transmutable data of the building through a model. We conclude that the five-layered architecture facilitates the exchange of transmutable data across diverse BMSs, making it a valuable tool for researchers and practitioners.

关键词： Temperature sensors Meters Architecture Buildings Time series analysis Data models Energy efficiency Sensors Stakeholders Building management systems

来源：评论

学校读者我要写书评

暂无评论

A Transformation Model for Different Granularity Linguistic Concept Formal Context 2nd

A Transformation Model for Different Granularity Linguistic ...

引用

2nd International Conference on Artificial Intelligence Logic and Applications, AILA 2022

作者： Kang, Ning Pang, Kuo Zou, Li Sun, Meiqiao Software Engineering Institute East China Normal University Shanghai200062 China Information Science and Technology College Dalian Maritime University Dalian116026 China School of Computer Science and Technology Shandong Jianzhu University Jinan250102 China School of Computer and Information Technology Liaoning Normal University Dalian116081 China

ISBN: (纸本)9789811975097

Due to certain differences in people's knowledge level and personal preferences, for the description of the same thing phenomenon, different people usually choose different granularity of linguistic term sets to make judgments, so it is necessary to propose a transformation method for linguistic concept formal context with different granularity. In this paper, we firstly define the normalized distance between multi-granularity linguistic formal contexts and linguistic terms for different granularity linguistic concept formal context. Then, the normalized distance from linguistic terms to intermediate linguistic terms is kept constant to realize the transformation for different granularity linguistic concept formal context. Finally, under different threshold constraints, we achieve dynamic transformation of formal contexts of linguistic values by adjusting the thresholds. The transformation method process is reversible and it can avoid information loss. © 2022, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.

关键词： Linguistics

来源：评论

学校读者我要写书评

暂无评论

Traffic Prediction Using Lightweight Large Model in UAV-Assisted Mobile Computing for Time-Critical Consumer Electronics

引用

IEEE Transactions on Consumer Electronics 2025年

作者： Jiang, Yunzhe Liu, Yinqiu Cui, Huajun Liu, Jinghuan Mao, Yingchi He, Xiaoming University of Electronic Science and Technology of China College of Communication and Information Engineering Chengdu China Nanyang Technological University School of Computer Science and Engineering Singapore PowerChina Beijing Engineering Corporation Limited Digital Intelligence Research Institute Beijing China Hohai University College of Computer Science and Software Engineering Nanjing China Nanjing University of Posts and Telecommunications College of Internet of Things Nanjing China

Unmanned Aerial Vehicles (UAVs) have emerged as an important category of consumer electronics, offering versatile applications and enhanced operational efficiency. In the context of Mobile Edge Computing (MEC), UAVs can improve task execution and service quality by providing dynamic allocation of computational and data processing resources at the network edge. However, one critical issue in UAV-assisted MEC is task traffic prediction, which directly impacts the efficiency of task offloading strategies. Inspired by the recent success of large models, we propose a novel Lightweight Large Model with Temporal-Spatial Attention (L-LMTSA), designed to maximize task traffic prediction accuracy and minimize the overall computation latency. Specifically, L-LMTSA integrates a Temporal Attention Model (TAM) and a Spatial Attention Model (SAM) to capture multi-dimensional information, such as task flow location and temporal characteristics. In addition, we develop a pre-trained lightweight GPT-2 model by employing a one-dimensional convolutional input encoding layer and fine-tuning its parameters for predictive accuracy. Additionally, we optimize the UAV's trajectory and task offloading strategy using a three-dimensional artificial potential field informed by the predicted task traffic. Extensive experiments demonstrate that L-LMTSA not only improves training speed and prediction accuracy but also significantly reduces computational latency, resulting in more efficient UAV-assisted MEC experiences for consumers. © 2024 IEEE.

关键词： Unmanned aerial vehicles (UAV)

来源：评论

学校读者我要写书评

暂无评论

R2DET: EXPLORING RELAXED ROTATION EQUIVARIANCE IN 2D OBJECT DETECTION

arXiv

引用

arXiv 2024年

作者： Wu, Zhiqiang Liu, Yingjie Dong, Hanlin Tang, Xuan Yang, Jian Jin, Bo Chen, Mingsong Wei, Xian Software Engineering Institute East China Normal University China School of Communication and Electronic Engineering East China Normal University China School of Geospatial Information Information Engineering University China School of Computer Science and Technology Tongji University China

Group Equivariant Convolution (GConv) empowers models to explore underlying symmetry in data, improving performance. However, real-world scenarios often deviate from ideal symmetric systems caused by physical permutation, characterized by non-trivial actions of a symmetry group, resulting in asymmetries that affect the outputs, a phenomenon known as Symmetry Breaking. Traditional GConv-based methods are constrained by rigid operational rules within group space, assuming data remains strictly symmetry after limited group transformations. This limitation makes it difficult to adapt to Symmetry-Breaking and nonrigid transformations. Motivated by this, we mainly focus on a common scenario: Rotational Symmetry-Breaking. By relaxing strict group transformations within Strict Rotation-Equivariant group Cn, we redefine a Relaxed Rotation-Equivariant group Rn and introduce a novel Relaxed Rotation-Equivariant GConv (R2GConv) with only a minimal increase of 4n parameters compared to GConv. Based on R2GConv, we propose a Relaxed Rotation-Equivariant Network (R2Net) as the backbone and develop a Relaxed Rotation-Equivariant Object Detector (R2Det) for 2D object detection. Experimental results demonstrate the effectiveness of the proposed R2GConv in natural image classification, and R2Det achieves excellent performance in 2D object detection with improved generalization capabilities and robustness. The code is available in https://***/wuer5/r2det. Copyright © 2024, The Authors. All rights reserved.

关键词： Object detection

来源：评论

学校读者我要写书评

暂无评论

Facial Emotion Recognition using Deep Learning (FERDL)

Facial Emotion Recognition using Deep Learning (FERDL)

引用

IEEE International Conference on Multi Topic

作者： Hafiz Arslan Ramzan Ahmed Sohaib Sadia Ramzan Tehmina Kalsum Muhammad Aqeel Mohsin Majeed Institute of Computer and Software Engineering Khwaja Fareed University of Engineering and Information Technology Rahim Yar Khan Pakistan Department of Computer Science National University of Modern Languages Multan Pakistan Department of Software Engineering University of Engineering & Technology Taxila Pakistan

Facial Emotion Recognition is one of the in-demand and rapidly growing research topics in the domain of computer Vision (CV) and artificial intelligence (AI). The ability to identify or detect human emotions from real-time facial expressions (FEs) has vast conceivable applications in different domains, such as sentiment analysis, human-computer interaction, human resource management, security, and human psychology. In this paper, a Convolutional Neural Network (CNN) based deep learning model is trained with haar-cascade classifier to recognize the real-time FEs. The suggested model is specially trained to categorize the FEs into one of the seven emotion categories, namely six basic emotions (sad, happy, angry, surprised, disgusted, fear) and a neutral emotion. It includes several convolutional layers, as well as fully connected neurons followed by max-pooling layers, and soft-max activation function with the corresponding seven classes. ReLU activation functions along with various kernels to enhance filtering depth, and extraction of facial features. FER-2013 dataset is used for experimentation purpose. To improve the classification performance and model accuracy, a data augmentation technique is used for rescaling and horizontal flipping. The proposed model outperforms the previous related works by achieving a validation accuracy of 71.96% and training accuracy above 90%, with fewer epochs.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Deep learning based DeepFake video detection

Deep learning based DeepFake video detection

引用

2023 International Conference on Smart Computing and Application, ICSCA 2023

作者： Guefrachi, Sarra Ben Jabra, Marwa Alsharabi, Naif A. Ben Othman, Mohamed Tahar Alharabi, Yasser O. Alkholidi, Abdulsalam Hammam, Habib Université de Moncton Faculty of Engineering MonctonNBE1A3E9 Canada Spectrum of Knowledge Production & Skills Development Sfax3027 Tunisia Computer Engineering University of Ha'il Computer Science Amran University Yemen Ha'il55476 Saudi Arabia College of Computer Qassim University Department of Computer Science Buraydah51452 Saudi Arabia College of Computer Sc. & Engneering Ha'il University Computer Science and Information Ha'il Saudi Arabia Canadian Institute of Technology College Department of Software Engineering Tirana Albania International Institute of Technology and Management Commune d'Akanda BP Libreville1989 Gabon School of Electrical Engineering University of Johannes-burg Dept of Electrical and Electronic Eng. Science Tirana Albania

ISBN: (纸本)9798350347050

Recent advances in DeepFake face-swapping technology have made it simple to create fake videos that appear remarkably real. Since it has been employed in numerous instances for deceit, extortion, and the falsification of facts, its widespread use has generated a huge social, security, and political risk. Its use on websites and social media has become more widespread. Detecting this crime is becoming more and more important due to the potential harm false videos may inflict on a global scale. This research offers a method for building a deep learning model that really can tell the difference between authentic and false videos. The article describes how to create new models based on the VGG16 neural network, a previously created neural network that does image categorization, using transfer learning in the computer vision field. Deep learning is still becoming better at both producing and spotting DeepFakes. DeepFake detection algorithms are developed using dated public datasets, and as a result, they may become obsolete with time. and require continual updating. The research findings are encouraging, and our results reached an accuracy rate of over 90%. © 2023 IEEE.

关键词： Crime

来源：评论

学校读者我要写书评

暂无评论

Deep Learning-Based E-Learning Solution for Identifying and Bridging the Knowledge Gap in Primary Education

Deep Learning-Based E-Learning Solution for Identifying and ...

引用

International Research Conference on Smart Computing and Systems engineering (SCSE)

作者： D.P.H Arunoda S.R Walpola S.M.I Piumira A.D.M.P Athukorala Thusithanjana Thilakarathna Sanjeevi Chandrasiri Department of Computer Science and Software Engineering Faculty of Computing Sri Lanka Institute of Information Technology Malabe Sri Lanka Department of Information Technology Faculty of Computing Sri Lanka Institute of Information Technology Malabe Sri Lanka

Educational teaching apps are primarily available in app stores to educate students in various contexts. Lack of educational resources, physical and mental health conditions, and poverty cause some students to skip school and move on to the next school grade without completing the course content of the previous grade. Most of the available apps focus on specific content to cover. The Smart Primary Education Tutor (SPET) teaching app specifically focuses on the missed content by analyzing their knowledge gap and providing lessons to cover the missed content. The main objective of SPET is to develop a methodology to identify the gap in student knowledge and fill the knowledge gap by teaching using smart techniques. SPET is determined to identify students’ interactions (attention, emotions) with the system to identify students’ ability to use the learning tool, identifying gaps in students’ knowledge levels compared to their actual grades using activities and voice-based technologies, teaching to cover the knowledge gap by providing engaging activities and lessons and evaluating students by conducting a final assessment and analyze students’ knowledge and performance obtained through the system. Students between the ages of 5 and 8 are targeted in the community to apply. The solution embeds deep learning-based models including attention classification models using head posture estimation, facial expression recognition, and eye gaze estimation, speech recognition models to identify provided verbal answers, handwriting recognition models to evaluate student performance, and smart teaching. The child emotion recognition model achieved 93% accuracy. The Attention span evaluation model achieved 85% accuracy. The handwritten numerical and English character data recognition model which detects answers for the final assessment paper achieved 85% percent of accuracy.

关键词：

来源：评论

学校读者我要写书评

暂无评论

DatawiseAgent: A Notebook-Centric LLM Agent Framework for Automated Data science

arXiv

引用

arXiv 2025年

作者： You, Ziming Zhang, Yumiao Xu, Dexuan Lou, Yiwei Yan, Yandong Wang, Wei Zhang, Huaming Huang, Yu National Engineering Research Center for Software Engineering Peking University China School of Software & Microelectronics Peking University China School of Computer Science Peking University China Xi’an Jiaotong University China Institute of Basic Theory of Chinese Medicine China Academy of Chinese Medical Sciences China

Data science tasks are multifaceted, dynamic, and often domain-specific. Existing LLM-based approaches largely concentrate on isolated phases, neglecting the interdependent nature of many data science tasks and limiting their capacity for comprehensive end-to-end support. We propose DatawiseAgent, a notebook-centric LLM agent framework that unifies interactions among user, agent and the computational environment through markdown and executable code cells, supporting flexible and adaptive automated data science. Built on a Finite State Transducer(FST), DatawiseAgent orchestrates four stages, including DSF-like planning, incremental execution, self-debugging, and post-filtering. Specifically, the DFS-like planning stage systematically explores the solution space, while incremental execution harnesses real-time feedback and accommodates LLMs’ limited capabilities to progressively complete tasks. The self-debugging and post-filtering modules further enhance reliability by diagnosing and correcting errors and pruning extraneous information. Extensive experiments on diverse tasks, including data analysis, visualization, and data modeling, show that DatawiseAgent consistently outperforms or matches state-of-the-art methods across multiple model settings. These results highlight its potential to generalize across data science scenarios and lay the groundwork for more efficient, fully automated workflows. Copyright © 2025, The Authors. All rights reserved.

关键词： Data science

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：