检索结果-内蒙古大学图书馆

International Journal of Mathematical, engineering and Management sciences 2025年第4期10卷 1032-1054页

作者： Purohit, Naresh Panwar, Subhash Department of Computer Science & Engineering Government Engineering College Bikaner Rajasthan Bikaner India

Since author’s writing styles are often ambiguous, writer recognition is an appealing research problem for handwritten manuscript investigation. Pattern identification allows for recognizing the author of a handwritten work. Due to the variety of text visuals, especially handwriting images, author recognition is challenging. Convolution Neural Network (CNN) excels in many fields. This paper presents a complete framework that emphasizes feature engineering and uses a simple, efficient deep neural network. Writer recognition begins with data collection. Programming a technique to generate several database texts enhances the data. To capture and categorize feature information, a multi-path CNN with a Bidirectional Long Short Term Memory (Bi-LSTM) module is developed. By integrating CNN and Bi-LSTMs, the proposed model combines spatial and temporal information, offering a comprehensive representation of handwriting. This synergy makes it particularly effective in handling the variability and complexity of offline writer recognition. Languages are occasionally blended when writing in multilingual regions. The system performed well on two publicly accessible handwritten English language benchmark databases, CVL and IAM, in addition to an in-house bilingual database comprising English and Hindi scripts. This study is unconventional since it addressed offline writer recognition using monolingual and bilingual handwritten text corpora. Analytically, our findings are encouraging compared to previous studies. The model achieves accuracies of 98.78% on the IAM dataset, 98.55% on the CVL dataset, and 99% on an in-house bilingual dataset. These results demonstrate a significant improvement over other state-of-the-art methodologies in offline writer recognition. Copyright © Ram Arti Publishers.

关键词： Feature Selection

来源：评论

学校读者我要写书评

暂无评论

3D brain image based tumor classification using ensemble of reinforcement transfer-based belief neural networks

引用

Multimedia Tools and Applications 2025年第14期84卷 14001-14028页

作者： Arora, Shraddha Lamba, Monika Department of Computer Science and Engineering The NorthCap University Haryana Gurugram India

The procedure of segmenting a brain tumour is an essential step in the field of medical image processing. In order to improve the efficacy of disease treatment choices and the likelihood of patient survival, the timely detection of a brain tumour is not only highly essential but a vital factor. For the purpose of cancer diagnosis, manually segmenting brain tumours from the large number of MRI images that are generated in clinical practice is a hard and time-consuming process. Automatic brain tumour image segmentation is a technique that absolutely has to exist. Using deep learning and machine learning architectures for segmentation and feature extraction, this research proposes a novel approach to the detection of brain tumours. In this case, the input consists of a collection of 3-D brain slice pictures that have been smoothed and with noise eliminated. After that, the picture that has been processed is segmented by using a convolutional radial function architecture that centres on contours. The segmented image features retrieval has been accomplished by the use of an ensemble of reinforcement transfer-based belief neural networks. The results of the experimental study have been completed with regard to precision, recall, accuracy, F-1 score, area under the curve, dice coefficient, and root mean square error. The strategy that was recommended was successful in achieving 97.57% accuracy, 97.15% precision, 89.37% recall, 83.56% F-1 score, 69.59% area under the curve, 63.83% dice coefficient, and 55.56% root mean square error. The sagittal, coronal, and axial planes of the pictures were shown in order to acquire the specific position of the tumour as well as its characteristics, which included its height, breadth, and depth. When it comes to the prediction of tumours, the findings of the assessment of semantic segmentation based on deep learning networks are highly positive. © The Author(s), under exclusive licence to Springer science+Business Media, LLC, part of Spri

关键词： Feature extraction

来源：评论

学校读者我要写书评

暂无评论

Image classification of lotus in Nong Han Chaloem Phrakiat Lotus Park using convolutional neural networks

引用

Artificial Intelligence in Agriculture 2024年第1期11卷 23-33页

作者： Thanawat Phattaraworamet Sawinee Sangsuriyun Phoempol Kutchomsri Susama Chokphoemphun Department of Electrical and Computer Engineering Faculty of Science and EngineeringKasetsart University Chalermphrakiat Sakonnakhon Province Campus SakonnakhonThailand Department of Computer and Information Science Faculty of Science and EngineeringKasetsart University Chalermphrakiat Sakonnakhon Province Campus SakonnakhonThailand

The Nong Han Chaloem Phrakiat Lotus Park is a tourist attraction and a source of learning regarding lotus ***,as a training area,it lacks appeal and learning motivation due to its conventional presentation of information regarding lotus *** current study introduced the concept of smart learning in this setting to increase interest and motivation for *** neural networks(CNNs)were used for the classification of lotus plant species,for use in the development of a mobile application to display details about each *** scope of the study was to classify 11 species of lotus plants using the proposed CNN model based on different techniques(augmentation,dropout,and L2)and hyper parameters(dropout and epoch number).The expected outcome was to obtain a high-performance CNN model with reduced total parameters compared to using three different pre-trained CNN models(Inception V3,VGG16,and VGG19)as *** performance of the model was presented in terms of accuracy,F1-score,precision,and recall *** results showed that the CNN model with the augmentation,dropout,and L2 techniques at a dropout value of 0.4 and an epoch number of 30 provided the highest testing accuracy of *** best proposed model was more accurate than the pre-trained CNN models,especially compared to Inception *** addition,the number of total parameters was reduced by approximately 1.80–2.19 *** findings demonstrated that the proposed model with a small number of total parameters had a satisfactory degree of classification accuracy.

关键词： Lotus Park Classification Smart learning park Convolutional neural network

来源：评论

学校读者我要写书评

暂无评论

Hybrid CNN-ViT architecture to exploit spatio-temporal feature for fire recognition trained through transfer learning

引用

Multimedia Tools and Applications 2025年第8期84卷 4703-4732页

作者： Shahid, Mohammad Wang, Hong-Cyuan Chen, Yung-Yao Hua, Kai-Lung Dept. of Computer Science and Information Engineering National Taiwan University of Science and Technology Taipei Taiwan Dept. of Electronic and Computer Engineering National Taiwan University of Science and Technology Taipei Taiwan

Fires are becoming one of the major natural hazards that threaten the ecology, economy, human life and even more worldwide. Therefore, early fire detection systems are crucial to prevent fires from spreading out of control and causing destruction. Based on vision sensors, many fire detection techniques have evolved with the recent surge of curiosity in deep learning, which exploits the spatial features of individual images. However, fire can take different forms, scales, and combustion materials can produce different colors, making accurate fire detection from an image challenging. Small fires captured from long-distance cameras lack salient features, further complicating detection. This paper proposes a hybrid structure that uses attention-enhanced convolutional neural networks and vision transformers (CNN-ViT) to detect fire. The proposed CNN-ViT first pays spatial attention to every frame and then aggregates temporal contextual information from neighboring frames to improve detection performance. Due to the limited availability of training fire datasets, the study employs deep transfer learning for feature extraction using pre-trained CNN. We used various metrics to examine the efficacy of the proposed approach. The results showed that the CNN-ViT method outperformed previous models based on spatial-temporal characteristics by achieving a relative improvement in accuracy and F1 score. The satisfactory results on images contaminated with different intensities of noise confirm the robustness of the approach. © The Author(s), under exclusive licence to Springer science+Business Media, LLC, part of Springer Nature 2024.

关键词： Fires

来源：评论

学校读者我要写书评

暂无评论

Smart traffic control: machine learning for dynamic road traffic management in urban environments

引用

Multimedia Tools and Applications 2025年第12期84卷 10321-10345页

作者： Khan, Hameed Thakur, Jitendra Singh Deptt of Computer Science & Engineering Jabalpur Engineering College Madhya Pradesh Jabalpur India

Roadside and outside environmental elements contribute to the road traffic setting's highly dynamic and turbulent nature. The human factor, primarily disregarded in the present research, is an essential element that contributes to the traffic context in addition to infrastructure-related elements like traffic signals, road infrastructure, and other road networks. Timing the green light and tracing the object that makes the incorrect turn using real-time visual information for traffic monitoring are still challenging tasks for the conventional traffic control system. We describe a self-adaptive real-time algorithm based on real-time traffic flow and monitoring. Combining image processing with AI-powered, self-adaptive machine learning for controlling traffic clearance at intersections is a forward-thinking approach with great potential. The suggested system uses the You Only Look Once v3 (YOLOv3) model and single image processing using a neural network to determine traffic clearance at the signal. YOLOv3 method to recognize objects from video frames. Subsequently, the centroid object tracking technique is used to monitor the movement of each vehicle within a proposed framework. We implemented algorithms to identify vehicles traveling in the incorrect direction based on their trajectories. This integrated approach enhances accurate object recognition, real-time vehicle tracking, and the detection of traffic violations, enhancing overall road safety measures. The experimental findings are quite promising, achieving an exclusive comparison between expected and actual vehicle numbers is crucial for any traffic monitoring system. The average object detection accuracy of 88.43% is impressive, and the exceptional 90.45% accuracy in tracking vehicles engaging in wrong turns or reckless driving behaviors is particularly noteworthy—it provides the system's ability to address safety concerns effectively. Integrating a Convolutional Neural Network (CNN) into the algorithm to all

关键词： Machine learning

来源：评论

学校读者我要写书评

暂无评论

Privacy-preserving intrusion detection with federated learning: enhancing privacy and efficiency in IoT networks

引用

International Journal of Machine Learning and Cybernetics 2025年 1-19页

作者： Sharma, Hemant Yadav, Gyan Singh Department of Computer Science and Engineering IIIT Kota Rajasthan Kota India

Intrusion detection systems (IDS) are crucial in the identification of unauthorized activities on a digital network, enabling cybersecurity measures to initiate prevention protocols to protect the security of their networks and data. With Internet of Things (IoT) as a ubiquitous phenomenon, IDS has become increasingly important owing to the massive growth of connected devices. Machine learning algorithms can detect anomalies in large datasets, but centralized methods pose security risks by gathering all data in one place. Federated learning (FL) offers a safer alternative, allowing multiple clients to collaborate on a model without sharing data, reducing the risk of data leaks and enhancing privacy. In this paper, we propose a FL-based IDS. Herein, we have applied the FL approach to a number of clients that are grouped in sets where the number of clients in a set varies from 2 to 10. After that, we applied feature ranking via a random forest approach, wherein the features are ranked based on their importance to find the optimal and reduced dataset for reducing the inference time. Further, three models for each of the clients were trained using the refined dataset obtained. Next, the central server aggregates the data from the clients using three methods. We have updated the dataset with two types of scenarios, one with two classes (Benign and Attacks) and the second with a multi-class (Benign, Dodag, Flooding, Rank, Blackhole) for identifying the attack. The experimental results establish that the proposed methodology achieves superior performance as compared to similar existing methodologies. © The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature 2025.

关键词： Federated learning

来源：评论

学校读者我要写书评

暂无评论

Robust subspace clustering via two-way manifold representation

引用

Multimedia Tools and Applications 2024年 1-18页

作者： Ezeora, Nnamdi Johnson Anichebe, Gregory Emeka Nzeh, Royransom Chiemela Uzo, Izuchukwu Uchenna Department of Computer Science University of Nigeria Nsukka Nigeria School of Computer Science and Communication Engineering JiangSu University JiangSu212013 China

Subspace clustering has shown great potential in discovering the hidden low-dimensional subspace structures in high-dimensional data. However, most existing methods still face the problem of noise distortion and overlapping subspaces. To tackle this problem and ensure that each sample is only assigned to a single subspace, a new method is proposed in this paper. Specifically, a two-way learning technique is introduced by inducing data manifold via two representative structures. The first is a low-rank structure learned directly from original data. The second structure is an affinity matrix obtained via a k symmetric nearest neighbor graph. By further introducing a dual regularization term, both structures are allowed to guild themselves adaptively to find robust clustering directly without spectral post-processing. In order to evaluate the effectiveness of the proposed method, several experiments are conducted on multiple benchmark datasets. The experimental results, evaluated using six standard metrics, clearly demonstrate that the proposed method outperforms state-of-the-art methods. © The Author(s), under exclusive licence to Springer science+Business Media, LLC, part of Springer Nature 2024.

关键词： Clustering algorithms

来源：评论

学校读者我要写书评

暂无评论

Multi objective binary Rao feature optimization for software defect prediction using machine learning models

引用

Soft Computing 2024年第23期28卷 13541-13565页

作者： Manchala, Pravali Tiwari, Ankur Bisi, Manjubala Computer Science and Engineering NIT Warangal Hanamkonda506004 India

Software defect prediction (SDP) is crucial for enhancing software quality and reliability, but class imbalance and irrelevant features hinder its performance, necessitating dataset optimization. In this work, MOBR first generates solutions using the Rao algorithm, and then a diverse set of Pareto optimal solutions are produced for decision-making. A multi-criteria decision-making technique called the weighted sum method selects the optimal feature set from the Pareto front. To evaluate the effectiveness of the MOBR approach, a comparison is conducted with Chi-Square (CS), Information Gain (IG), Single Objective Binary Rao (BR), and Multi-objective NSGA-II over twenty-four benchmark Promise and NASA datasets with imbalanced and balanced data using Decision Tree (DT), K Nearest Neighbour (KNN), Naive Bayes (NB), and Support Vector Machine (SVM) prediction models. Our MOBR model is evaluated using five assessment measures and statistically compared, further win-draw-loss comparison performed to compare each prediction model. The experimental and statistical analysis demonstrates that In terms of FOR, Recall, F1-Score, AUC, and G-mean, MOBR gets top ranks of 3.5, 1, 5, 2.5, and 1 over unbalanced data and 3, 1.5, 1.5, 6, 2.5 over balanced data, respectively. Moreover, MOBR with DT gives superior performance. Hence, it can be concluded that the MOBR can effectively improve model performance while minimizing the number of features. © The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature 2024.

关键词： Prediction models

来源：评论

学校读者我要写书评

暂无评论

A TSN-Like Slot-Based Scheduler for Improved Wireless Quality and Platoon Formation in Smart Factories

IEEE Open Journal of the Communications Society

引用

IEEE Open Journal of the Communications Society 2025年 6卷 3003-3013页

作者： Feraudo, Angelo Garbugli, Andrea Bellavista, Paolo University of Bologna Department of Computer Science and Engineering Bologna40126 Italy

The platooning of mobile robots, facilitated by Device-to-Device (D2D) communications, has become central in Industry 4.0, enhancing material transport, reducing energy consumption, and improving safety in smart factories. However, as the number of participating robots increases, maintaining effective communication and coordination becomes increasingly challenging. Shared spectrum usage in D2D networks can lead to communication collisions, particularly in highly dynamic scenarios, posing significant challenges in industrial environments. This paper introduces a novel slot-based solution inspired by Time Sensitive Networking (TSN) to address these challenges, ensuring reliable, low-latency communication while optimizing resource allocation and communication efficiency in platooning systems. To achieve this, we propose a controller named TSNCtl, operating at the application layer of the network stack. TSNCtl leverages a finite state machine (FSM) to manage platoon formation and employs slot-based scheduling for efficient message dissemination. Using the OMNeT++ simulation framework and INET library, we demonstrate the effectiveness of TSNCtl in reducing packet collisions across a variety of scenarios. Our experimental results highlight a significant improvement over the traditional CSMA/CA baseline employed in Wi-Fi-based protocols such as IEEE 802.11p. For instance, TSNCtl achieves near-zero packet collisions with appropriate configurations, even in highly dynamic environments. © 2020 IEEE.

关键词： 5G mobile communication systems

来源：评论

学校读者我要写书评

暂无评论

Speech signal’s phase information based Alzheimer’s disease detection using deep learning

引用

International Journal of Speech Technology 2025年 1-14页

作者： Kumar, Mohit Sushant Yadav, Arun Kumar Department of Computer Science & Engineering NIT Hamirpur India

Alzheimer’s disease (AD) is a debilitating neurodegenerative disorder that affects millions worldwide. Early and accurate diagnosis is crucial for timely intervention and management, as it can significantly improve patient outcomes and quality of life. This study explores a novel deep learning approach for AD detection using speech analysis, a promising non-invasive technique. We analysed the performance of various feature types, including magnitude spectrograms, raw phase, modulated group delay (MGD), and instantaneous frequency (IF) phase, on custom Convolutional Neural Network (CNN) and pre-trained ResNet50 models. The findings demonstrate that the magnitude spectrogram achieved the highest accuracy among single-input models. However, a dual-input model that combines magnitude and instantaneous frequency information from pre-trained ResNet50 models achieved superior performance, reaching an accuracy of 90%, precision of 95%, recall of 90%, and F1-score of 92.4%. This suggests that integrating magnitude and phase features effectively captures the nuances of speech patterns, leading to improved classification for AD detection. Moreover, the proposed approach outperforms previous state-of-the-art methods, highlighting its potential as a valuable tool for early AD diagnosis. © The Author(s), under exclusive licence to Springer science+Business Media, LLC, part of Springer Nature 2025.

关键词： Speech analysis

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：