检索结果-内蒙古大学图书馆

IAENG International Journal of computer science 2025年第1期52卷 23-31页

作者： Guo, Ying Wang, Li School of Computer Science and software Engineering University of Science and Technology Liaoning Anshan114051 China College of Computer Science and Technology Liaoning Anshan114051 China

End-to-end training has emerged as a prominent trend in speech recognition, with Conformer models effectively integrating Transformer and CNN architectures. However, their complexity and high computational cost pose deployment challenges. To address these issues, we propose a multi-task Chinese speech recognition method based on the Squeezeformer model. We replace the FMCF structure in Conformer with an MF/CF structure, leveraging the convolutional module as a local Multi- Head Attention (MHA) module to enhance efficiency. Multilevel down-sampling and up-sampling using a time-series U-Net further reduce computational costs. By eliminating redundant LayerNorm layers and employing depthwise separable convolutions, we streamline the model, reduce parameters, and lower deployment costs. An Adaptor Layer is integrated into the MHSA module to mitigate the vanishing gradient problem, and a ScaleVar Layer is added to enhance flexibility. Additionally, the RealFormer module is introduced on the decoding side to improve context understanding. Combining Connectionist Temporal Classification (CTC) with attention-based encoding and decoding models for multi-task learning improves performance and accuracy. Experimental results show that the proposed method reduces the parameters on AISHELL-1 dataset by 16% and reduces the character error rate to 5.50%. At the same time, it also shows good performance on AISHELL-2 dataset. © (2025), (International Association of Engineers). All rights reserved.

关键词： Speech recognition

来源：评论

学校读者我要写书评

暂无评论

DMF: A Deep Multimodal Fusion-Based Network Traffic Classification Model

引用

computers, Materials & Continua 2025年第5期83卷 2267-2285页

作者： Xiangbin Wang Qingjun Yuan Weina Niu Qianwei Meng Yongjuan Wang Chunxiang Gu Henan Key Laboratory of Network Cryptography Technology Information Engineering UniversityZhengzhou450001China School of Computer Science and Engineering University of Electronic Science and Technology of ChinaChengdu611731China

With the rise of encrypted traffic,traditional network analysis methods have become less effective,leading to a shift towards deep learning-based *** these,multimodal learning-based classification methods have gained attention due to their ability to leverage diverse feature sets from encrypted traffic,improving classification ***,existing research predominantly relies on late fusion techniques,which hinder the full utilization of deep features within the *** address this limitation,we propose a novel multimodal encrypted traffic classification model that synchronizes modality fusion with multiscale feature ***,our approach performs real-time fusion of modalities at each stage of feature extraction,enhancing feature representation at each level and preserving inter-level correlations for more effective *** continuous fusion strategy improves the model’s ability to detect subtle variations in encrypted traffic,while boosting its robustness and adaptability to evolving network *** results on two real-world encrypted traffic datasets demonstrate that our method achieves a classification accuracy of 98.23% and 97.63%,outperforming existing multimodal learning-based methods.

关键词： Deep fusion intrusion detection multimodal learning network traffic classification

来源：评论

学校读者我要写书评

暂无评论

RoleNet: A multiple features fusion network for role classification in cantonese opera

引用

Multimedia Tools and Applications 2025年 1-14页

作者： Li, Yue Peng, Zhengwei Xu, Di Chen, Yuanguang Chen, Guoan School of Computer Science and Engineering South China University of Technology Guangdong Guangzhou510006 China School of Computer Science and Engineering Sun Yat-sen University Guangdong Guangzhou510006 China

Cantonese opera, a key facet of Chinese traditional opera, boasts profound cultural and artistic value and has been designated as intangible cultural heritage. The use of certain roles is a basic concept in Cantonese opera, where each role has a specific style of singing, movement, and costume that performers are trained to perform throughout their careers. Therefore, identifying the role category of characters in a play can provide theoretical and systematic foundations for further researches and artistic explorations. By dissecting musical traits and performance styles of each role, comprehensive studies on its regional and artistic nuances are enabled. To achieve role classification in Cantonese opera, we propose RoleNet, an integration network that consists of SincNets, transformers, and a feature fusion block. For a given musical fragment (e.g., an audio signal), SincNets extract 1D features at multiple scales and transformers extract features from time and frequency axes from the 2D Mel Spectrogram. Subsequently, the extracted features are concatenated by the fusioner using multiple features selection strategy to perform role classification tasks. The experimental results on a real-world dataset demonstrated the superior performance of RoleNet compared to single-objective methods. An overall classification accuracy of 98.67% was achieved. The codes are available at https://***/AaronPeng920/RoleNet. © The Author(s), under exclusive licence to Springer science+Business Media, LLC, part of Springer Nature 2025.

关键词： Feature Selection

来源：评论

学校读者我要写书评

暂无评论

Deep Convolutional Network Based Machine Intelligence Model for Satellite Cloud Image Classification

引用

Big Data Mining and Analytics 2023年第1期6卷 32-43页

作者： Kalyan Kumar Jena Sourav Kumar Bhoi Soumya Ranjan Nayak Ranjit Panigrahi Akash Kumar Bhoi Department of Computer Science and Engineering Parala Maharaja Engineering CollegeBerhampur 761003India Amity School of Engineering and Technology Amity UniversityUttar Pradesh 201303India Department of Computer Applications Sikkim Manipal Institute of TechnologySikkim Manipal UniversitySikkim 737102India Directorate of Research Sikkim Manipal UniversityGangtokSikkim 737102India

As a huge number of satellites revolve around the earth,a great probability exists to observe and determine the change phenomena on the earth through the analysis of satellite images on a real-time ***,classifying satellite images plays strong assistance in remote sensing communities for predicting tropical *** this article,a classification approach is proposed using Deep Convolutional Neural Network(DCNN),comprising numerous layers,which extract the features through a downsampling process for classifying satellite cloud *** is trained marvelously on cloud images with an impressive amount of prediction *** time decreases for testing images,whereas prediction accuracy increases using an appropriate deep convolutional network with a huge number of training dataset *** satellite images are taken from the Meteorological&Oceanographic Satellite Data Archival Centre,the organization is responsible for availing satellite cloud images of India and its *** proposed cloud image classification shows 94% prediction accuracy with the DCNN framework.

关键词： satellite images satellite image classification cyclone prediction Deep Convolutional Neural Network(DCNN) features layers down-sampling process

来源：评论

学校读者我要写书评

暂无评论

Survey on One-Shot Learning for Image Recognition using Machine Learning and Deep Learning Techniques 5

Survey on One-Shot Learning for Image Recognition using Mach...

引用

5th International Conference on Smart Electronics and Communication, ICOSEC 2024

作者： Kadam, Ganesh Vitthal Dubey, Ghanshyam Prasad Computer Science Amity School of Engineering & Technology Gwalior Maharajpura Madhya Pradesh Gwalior474005 India Amity School of Engineering & Technology ASET-CSE Maharajpura Madhya Pradesh Gwalior474005 India

ISBN: (纸本)9798331504403

Despite their success, the traditional deep neural networks are unable to rapidly generalize from a limited number of samples. To address the issue of large training data requirements with deep neural networks, one-shot learning restricting the number of data in each category is developed to attain enhanced results. Specifically, the traditional neural networks lead to over-fitting problems in training with a limited number of samples. In contrast, Machine Learning algorithms are capable of one-shot learning from limited samples like humans have the ability of oneshot learning. Hence, it is necessary to analyze one-shot object recognition techniques prevalent in image processing systems. This study will explore different one-shot learning techniques and methods from the state-of-the-art models and examine the merits and challenges of the approach with 25 literature papers. The effectiveness of the existing approaches will be evaluated by various techniques, datasets, and performance metrics are discussed in detail. At last, the paper discusses the potential future studies and challenges in achieving better detection accuracy. © 2024 IEEE.

关键词： Zero-shot learning

来源：评论

学校读者我要写书评

暂无评论

Biosignal Contrastive Representation Learning for Emotion Recognition of Game Users

引用

IEEE Transactions on Games 2024年第2期17卷 1-15页

作者： Li, Rongyang Ding, Jianguo Ning, Huansheng School of Computer and Communication Engineering University of Science and Technology Beijing China Department of Computer Science Blekinge Institute of Technology Karlskrona Sweden

Biosignal representation learning (BRL) plays a crucial role in emotion recognition for game users (ERGU). Unsupervised BRL has garnered attention considering the difficulty in obtaining ground truth emotion labels from game users. However, unsupervised BRL in ERGU faces challenges, including overfitting caused by limited data and performance degradation due to unbalanced sample distributions. Faced with the above challenges, we propose a novel method of biosignal contrastive representation learning (BCRL) for ERGU, which not only serves as a unified representation learning approach applicable to various modalities of biosignals but also derives generalized biosignals representations suitable for different downstream tasks. Specifically, we solve the overfitting by introducing perturbations at the embedding layer based on the projected gradient descent (PGD) adversarial attacks and develop the sample balancing strategy (SBS) to mitigate the negative impact of the unbalanced sample on the performance. Further, we have conducted comprehensive validation experiments on the public dataset, yielding the following key observations: 1) BCRL outperforms all other methods, achieving average accuracies of 76.67%, 71.83%, and 63.58% in 1D-2C Valence, 1D-2C Arousal and 2D-4C Valence/Arousal, respectively;2) The ablation study shows that both the PGD module (+7.58% in accuracy on average) and the SBS module (+14.60% in accuracy on average) have a positive effect on the performance of different classifications;3) BCRL model exhibits the certain generalization ability across the different games, subjects and classifiers. IEEE

关键词： Emotion Recognition

来源：评论

学校读者我要写书评

暂无评论

Privacy-preserving cross-domain recommendation using hybrid federated transfer learning

引用

Multimedia Tools and Applications 2025年第18期84卷 19343-19377页

作者： Goyal, Samta Jain Goyal, Rajeev Singh, Vinay Kumar Arunachalam, Rajesh Tripathi, Kuldeep Narayan Computer Science and Engineering ASET Amity University Madhya Pradesh Madhya Pradesh Gwalior India Electronics and Communication Engineering Saveetha School of Engineering Saveetha Institute of Medical and Technical Sciences Tamilnadu Thandalam Chennai India Madhya Pradesh Gwalior India School of Computer Science and Engineering Lovely Professional University Punjab Phagwara India

Cross-domain recommender systems, which serve as a panacea to data sparsity and cold-start problems, are increasingly vital for providing personalized content across diverse application domains like medicine, e-commerce, etc. This is accomplished by leveraging adequate ratings ad user profiles in one domain to ensure precise recommendations in another domain. However, the domains with good ratings are unwilling to share user ratings with third-party recommender systems/domains due to privacy-concern. It demands the need for privacy mechanism to transfer knowledge between domains securely. So, this research introduces the CrossRec-Hybrid Federated Transfer Learning (CrossRec-HFTL) framework to address cold start and privacy problems in a cross-domain recommendation system, which combines adversarial domain adaptation and federated learning to leverage user and item interactions from a labeled source domain to generate recommendations in an unlabeled target domain. CrossRec-HFTL uses a model architecture called CrossRec-CDAN, which consists of multiple generators and discriminators for domain adaptation. The generators produce user, item, and interaction embeddings, while the discriminators distinguish between source and target domain samples. Federated learning enables collaborative model training across multiple clients while maintaining data locality, further bolstered by encryption techniques for privacy preservation. The framework was evaluated on the Amazon dataset across three domain pairs: Home and kitchen to Office Products (HK → OP), CDs and vinyl to Digital Music (CDs → DM) and Digital Music to Music Instruments (DM → MI). The simulation findings reveal that CrossRec-HFTL outperforms baseline and modern recommendation models in terms of recommendation accuracy, achieving a very less RMSE of 0.82 and MAE of 0.58. This shows the robust performance of CrossRec-HFTL in cold-start scenarios and enhanced privacy preservation. © The Author(s), under exclusive lice

关键词： Recommender systems

来源：评论

学校读者我要写书评

暂无评论

Anomaly Detectionin Network Traffic Scenarios by Resampling and Majority Voting with Concept Drift: A Hybrid Approach 11

Anomaly Detectionin Network Traffic Scenarios by Resampling ...

引用

11th International Conference on Reliability, Infocom Technologies and Optimization, ICRITO 2024

作者： Singh, Richa Srivastava, Nidhi Kumar, Ashwani Amity Institute of Information Technology Amity University Lucknow Campus Lucknow India Srm Institute of Science & Technology Computer Science and Engineering Faculty of Engineering & Technology Ghaziabad India

ISBN: (纸本)9798350350357

Addressing data imbalance is a critical difficulty in the setting of a heavily skewed dataset because the dominant class disproportionately influences classifier accuracy, especially when inadequate data impede the learning of minority class characteristics. For multi-class cases, traditional binary classification approaches are inadequate. The suggested architecture uses a Weighted Majority Voting Classifier (WMVC), noise cleaning, limited under-sampling, and oversampling to overcome this. One important step is to set an average class size limit by dividing the entire sample number by the number of classes. One or more majority classes receive Random Under sampling, all minority classes receive an up sample using the Adaptive Synthetic Sampling Approach, and noise reduction is achieved via Tomek Link. The resulting WMVC is contrasted with Convolutional Neural Network (CNN) classifiers, XGBoost, and six additional ensemble techniques by utilizing tree-based ensemble algorithms. Comparative study shows that balanced data performs better than unbalanced data. WMVC significantly beats CNN, XGBoost, and the other ensemble methods, and it improves performance for difficult minority classes while successfully decreasing classifier bias towards the majority class. © 2024 IEEE.

关键词： Anomaly detection

来源：评论

学校读者我要写书评

暂无评论

Performance evaluation for multi-target tracking with temporal dimension specifics

引用

Chinese Journal of Aeronautics 2024年第2期37卷 446-458页

作者： Zhenzhen SU Hongbing JI Cong TIAN Yongquan ZHANG School of Computer Science and Technology Xidian UniversityXi’an 710071China School of Electronic Engineering Xidian UniversityXi’an 710071China

With the great development of Multi-Target Tracking(MTT)technologies,many MTT algorithms have been proposed with their own advantages and *** to the fact that requirements to MTT algorithms vary from the application scenarios,performance evaluation is significant to select an appropriate MTT algorithm for the specific application *** this paper,we propose a performance evaluation method on the sets of trajectories with temporal dimension specifics to compare the estimated trajectories with the true *** proposed method evaluates the estimate results of an MTT algorithm in terms of tracking accuracy,continuity and ***,its computation is based on a multi-dimensional assignment problem,which is formulated as a computable form using linear *** enhance the influence of recent estimated states of the trajectories in the evaluation,an attention function is used to reweight the trajectory errors at different time ***,simulation results show that the proposed performance evaluation method is able to evaluate many aspects of the MTT *** evaluations are worthy for selecting suitable MTT algorithms in different application scenarios.

关键词： Multi-target tracking Temporal dimension specifics Performance evaluation Random finite sets Linear programming

来源：评论

学校读者我要写书评

暂无评论

Blockchain - A Secure and Transparent Solution to Detect Counterfeit Products 1

引用

6th International Conference on Applied Machine Learning and Data Analytics, AMLDA 2023

作者： Tyagi, Ishaan Gupta, Rajat Upadhyay, Divya Dubey, Ashwani Kumar Department of Computer Science and Engineering ABES Engineering College UP Ghaziabad India Department of Electronics and Communication Engineering Amity School of Engineering and Technology Amity University Uttar Pradesh UP Noida India

ISBN: (数字)9783031554865

ISBN: (纸本)9783031554858

Counterfeit products have had a massive impact on manufacturing industries in the past couple of years. This is affecting the company’s name, sales, and profit. Blockchain innovation has acquired an interest in the course recently. The most important issue about this is currency exchange, but its application is not restricted to only Digital currency. This technology has the potential to influence different business sectors. Blockchain has brought high transparency and ease in how transactions are dealt with. Blockchain technology can be used to identify real products from fake ones. Using this technology, customers or users need not rely on third-party services for the product’s safety. The proposed system’s Quick Response (QR) code provides a robust technique to stop counterfeiting the products. Fake products can be detected using a Quick Response scanner, where a QR code is attached. This QR code is linked to the Blockchain network to identify the fake product. A block will be created to store the product-related information and its QR code, like product details, and generate a unique code for each block in the blockchain database. The user uploads and then compares the unique code of the product he intends to purchase with that stored in the blockchain network. If the code matches the code that the manufacturer generated, it will notify the customer, saying the QR code is matched;otherwise, it will notify the customer that the QR code is not matched and the product is fake. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024.

关键词： Smart contract

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：