检索结果-内蒙古大学图书馆

39th Annual AAAI Conference on Artificial Intelligence, AAAI 2025

作者： Wang, Mei Zhu, Xiaojie Hu, Ruimin Zhu, Dongliang Liao, Liang Ye, Mang National Engineering Research Center for Multimedia Software School of Computer Science Wuhan University China Hubei Key Laboratory of Multimedia and Network Communication Engineering Wuhan University China School of Cyber Science and Engineering Wuhan University Whuhan China Cyberspace Security Laboratory School of Network and Information Security Xidian University Xi'an China The School of Computer Science and Engineering Nanyang Technological University Singapore

ISBN: (纸本)157735897X

Fatigue is a critical factor contributing to accidents in industries such as safety monitoring and engineering construction. Fatigue exhibits dynamic complexity and non-stationary characteristics, so there are many intermediate states of short-term variation between alert and fatigue. Capturing and learning the signs of these intermediate states is essential for accurate fatigue assessment. However, current fatigue detection methods primarily rely on coarse-grained labels, typically spanning minutes to hours, and commonly treat alert and fatigue as two distinctly separate distributions, overlooking the expression of intermediate states and oversimplifying the rich distribution information of fatigue types and levels, thereby limiting detection effectiveness. To address these, this paper explores a refined representation of fatigue in terms of three dimensions: time, type, and level, and proposes a MultiDimensional Fine-Grained Modeling for Fatigue Detection (MDFG). This introduces the SmallLoss to extract trustworthy samples, utilizes clustering to identify diverse subtypes under alert and fatigued states, and establishes base class sets in each state. Subsequently, a complete base class set containing intermediate state bases is constructed using the base class synthesis method, which achieves the expression of intermediate fatigue states from absence to presence. Finally, fatigue levels are quantified based on the matching between samples and the complete base class set. Moreover, to cope with the complex variability of fatigue states, MDFG employs meta-learning for training. MDFG achieves an Average accuracy improvement of 10.0% and 12.1% on two real datasets compared to methods that do not consider fine-grained information. Extensive experiments demonstrate that the MDFG exhibits superior robustness and stability among current fatigue detection methods. Copyright © 2025, Association for the Advancement of Artificial Intelligence (***). All rights reserv

关键词： Safety factor

来源：评论

学校读者我要写书评

暂无评论

Self-Driven Entropy Aggregation for Byzantine-Robust Heterogeneous Federated Learning 41

Self-Driven Entropy Aggregation for Byzantine-Robust Heterog...

引用

41st International Conference on Machine Learning, ICML 2024

作者： Huang, Wenke Shi, Zekun Ye, Mang Li, He Du, Bo National Engineering Research Center for Multimedia Software School of Computer Science Wuhan University Wuhan China Taikang Center for Life and Medical Sciences Wuhan University Wuhan China

Federated learning presents massive potential for privacy-friendly collaboration. However, federated learning is deeply threatened by byzantine attacks, where malicious clients deliberately upload crafted vicious updates. While various robust aggregations have been proposed to defend against such attacks, they are subject to certain assumptions: homogeneous private data and related proxy datasets. To address these limitations, we propose Self-Driven Entropy Aggregation (SDEA), which leverages the random public dataset to conduct Byzantine-robust aggregation in heterogeneous federated learning. For Byzantine attackers, we observe that benign ones typically present more confident (sharper) predictions than evils on the public dataset. Thus, we highlight benign clients by introducing learnable aggregation weight to minimize the instance-prediction entropy of the global model on the random public dataset. Besides, with inherent data heterogeneity, we reveal that it brings heterogeneous sharpness. Specifically, clients are optimized under distinct distribution and thus present fruitful predictive preferences. The learnable aggregation weight blindly allocates high attention to limited ones for sharper predictions, resulting in a biased global model. To alleviate this problem, we encourage the global model to offer diverse predictions via batch-prediction entropy maximization and conduct clustering to equally divide honest weights to accommodate different tendencies. This endows SDEA to detect Byzantine attackers in heterogeneous federated learning. Empirical results demonstrate the effectiveness. Copyright 2024 by the author(s)

关键词： Federated learning

来源：评论

学校读者我要写书评

暂无评论

FreqSense: Universal and Low-Latency Adversarial Example Detection for Speaker Recognition with Interpretability in Frequency Domain

FreqSense: Universal and Low-Latency Adversarial Example Det...

引用

2025 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2025

作者： Huang, Yihuan Li, Yuanzhe Ren, Yanzhen Tu, Weiping Yang, Yuhong Key Laboratory of Aerospace Information Security and Trusted Computing Ministry of Education China School of Cyber Science and Engineering Wuhan University China National Engineering Research Center for Multimedia Software School of Computer Science Wuhan University China

ISBN: (纸本)9798350368741

Speaker recognition (SR) systems are particularly vulnerable to adversarial example (AE) attacks. To mitigate these attacks, AE detection systems are typically integrated into SR systems. To overcome the limitations of low detection accuracy, poor generalization, and high latency in existing schemes, this paper proposes FreqSense, an AE detection scheme based on frequency distribution features. FreqSense detects a variety of unknown AE attacks with low latency, and provides interpretability in its detection process. The basic idea of FreqSense is that AE typically introduce carefully designed noise in specific frequency bands that are associated with highly distinctive speaker identities. Therefore, leveraging the distributional variations in these frequency bands can effectively distinguish between AE and benign audio. FreqSense models frequency distribution features by integrating time-frequency transformation technology with a self-attention mechanism and employs a neural network-based classifier to distinguish between AE and benign audio. Experimental results show that FreqSense achieves an overall detection accuracy of 99.2%, surpassing state-of-the-art (SOTA) schemes by 27.2%. When confronting unknown AE attacks, FreqSense achieves a detection accuracy of 98.3% with a latency of just 0.0014 seconds. © 2025 IEEE.

关键词： Adversarial Attack Adversarial Example Detection Speaker Recognition

来源：评论

学校读者我要写书评

暂无评论

GoMatching: A Simple Baseline for Video Text Spotting via Long and Short Term Matching 38

GoMatching: A Simple Baseline for Video Text Spotting via Lo...

引用

38th Conference on Neural Information Processing Systems, NeurIPS 2024

作者： He, Haibin Ye, Maoyuan Zhang, Jing Liu, Juhua Du, Bo Tao, Dacheng School of Computer Science National Engineering Research Center for Multimedia Software Institute of Artificial Intelligence Wuhan University China College of Computing & Data Science Nanyang Technological University Singapore

Beyond the text detection and recognition tasks in image text spotting, video text spotting presents an augmented challenge with the inclusion of tracking. While advanced end-to-end trainable methods have shown commendable performance, the pursuit of multi-task optimization may pose the risk of producing sub-optimal outcomes for individual tasks. In this paper, we identify a main bottleneck in the state-of-the-art video text spotter: the limited recognition capability. In response to this issue, we propose to efficiently turn an off-the-shelf query-based image text spotter into a specialist on video and present a simple baseline termed GoMatching, which focuses the training efforts on tracking while maintaining strong recognition performance. To adapt the image text spotter to video datasets, we add a rescoring head to rescore each detected instance's confidence via efficient tuning, leading to a better tracking candidate pool. Additionally, we design a long-short term matching module, termed LST-Matcher, to enhance the spotter's tracking capability by integrating both long- and short-term matching results via Transformer. Based on the above simple designs, GoMatching delivers new records on ICDAR15-video, DSText, BOVText, and our proposed novel test set with arbitrary-shaped text termed ArTVideo, which demonstrates GoMatching's capability to accommodate general, dense, small, arbitrary-shaped, Chinese and English text scenarios while saving considerable training budgets. © 2024 Neural information processing systems foundation. All rights reserved.

关键词：

来源：评论

学校读者我要写书评

暂无评论

A Hybrid Model of Multiple CNNs with Attention Mechanism for Optimized Facial Expression Recognition 5

A Hybrid Model of Multiple CNNs with Attention Mechanism for...

引用

5th International Conference on Artificial Intelligence and Computer engineering, ICAICE 2024

作者： Chen, Wanqiu Sun, Yuting Yang, Yihao Yu, Jiaxu Fuzhou University Multimedia Mobile and Web Development Fuzhou China Data Science and Big Data Technology Fuzhou University Fuzhou China Fuzhou University Software Engineering Fuzhou China

ISBN: (纸本)9798331528911

Facial expression recognition is integral to enhancing human-machine interactions, applicable across security, advertising, and social media. Traditionally, these systems relied on handcrafted features or basic algorithms, but advancements in computational power and neural network architecture like Convolutional Neural Networks (CNNs) have shifted this trend. This study presents a hybrid model combining multiple CNNs-MobileNetV1, MobileNetV2, and ResNet50-with an attention mechanism to improve the robustness and accuracy of facial expression recognition. Utilizing the strengths of these networks, the model utilizes an ensemble approach where feature extractions are integrated and optimized using a random forest to determine the most effective weightings for final predictions. This method not only harnesses the power of deep learning to enhance feature detection across diverse conditions but also implements transfer learning (i.e. ImageNet weights) to address challenges posed by small or non-varied datasets, significantly advancing the abilities of facial expression recognition systems. © 2024 IEEE.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

Mandarin Lombard Grid: A Noise Induced Lombard-Grid-Like Corpus of Standard Chinese 23

Mandarin Lombard Grid: A Noise Induced Lombard-Grid-Like Cor...

引用

23rd Annual Conference of the International Speech Communication Association, INTERSPEECH 2022

作者： Yang, Yuhong Chen, Xufeng Liu, Qingmu Tu, Weiping Chen, Hongyang Cai, Linjun National Engineering Research Center for Multimedia Software School of Computer Science Wuhan University Wuhan China Hubei Key Laboratory of Multimedia and Network Communication Engineering Wuhan China

The Lombard effect is natural, whereby speakers automatically adjust the vocal effort to facilitate speech understanding in noise. Since real-world applications are generally involved in noisy environments, the Lombard effect of highly variable speech features due to changing background noise is one of those challenges to match these real scenarios. Existing Lombard corpora show variations in the background noise level, ranging from 35 to 96 dB sound pressure level (SPL). However, it remains unclear if we need to collect all SPLs to build a comprehensive Lombard corpus. And most existing Lombard corpora are built for English;however, Mandarin and English are different in pronunciation. This paper describes our effort to build the first open-source Lombard corpus of standard Chinese, the Mandarin Lombard Grid. The effort involves three steps: (1) Classify Mandarin Lombard styles according to different background noise levels. (2) Create the corpus containing each style. (3) Analyze Mandarin Lombard effects showing their differences from English. We found three critical Lombard styles ranging from 30 dB to 85 dB-SPL and built the corpus containing the three Lombard styles and one reference plain style. Lombard effect analyses on this corpus showed consistency and some differences with the English Lombard Grid corpus. Copyright © 2022 ISCA.

关键词： Speech communication

来源：评论

学校读者我要写书评

暂无评论

PMMSD: Development of the Matrix Sentence Intelligibility Dataset for Mandarin with Lombard Effect 48

PMMSD: Development of the Matrix Sentence Intelligibility Da...

引用

48th IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP 2023

作者： Pei, Hanchen Yang, Yuhong Chen, Xufeng Liu, Qingmu Chen, Hongyang Tu, Weiping Lin, Song Wuhan University National Engineering Research Center for Multimedia Software School of Computer Science Wuhan China Guangdong Oppo Mobile Telecommunications Corporation China

ISBN: (纸本)9781728163277

This paper presents a Paired Mandarin Matrix Sentence Dataset (PMMSD), which will be available after publication. PMMSD is the first Mandarin matrix sentence intelligibility dataset containing both plain and Lombard speech for scientific research. The results verify that different Lombard styles would affect word intelligibility to different degrees and the Lombard effect helps maintain homogeneous intelligibility against contextual interference. All of the discoveries indicate that the Lombard effect should be considered when building intelligibility datasets with noise in the future. © 2023 IEEE.

关键词： intelligibility Lombard effect Matrix sentences SRT test

来源：评论

学校读者我要写书评

暂无评论

Mutual Retinex: Combining Transformer and CNN for Image Enhancement

引用

IEEE Transactions on Emerging Topics in Computational Intelligence 2024年第3期8卷 2240-2252页

作者： Jiang, Kui Wang, Qiong An, Zhaoyi Wang, Zheng Zhang, Cong Lin, Chia-Wen Harbin Institute of Technology School of Computer Science and Technology Harbin150001 China Wuhan University National Engineering Research Center for Multimedia Software School of Computer Science Wuhan430072 China Wuhan Polytechnic University School of Mathematics and Computer Science Wuhan430024 China National Tsing Hua University Department of Electrical Engineering Institute of Communications Engineering Hsinchu30013 Taiwan

Images captured in low-light or underwater environments are often accompanied by significant degradation, which can negatively impact the quality and performance of downstream tasks. While convolutional neural networks (CNNs) and Transformer architectures have made significant progress in computer vision tasks, there are few efforts to harmonize them into a more concise framework for enhancing such images. To this end, this study proposes to aggregate the individual capability of self-attention (SA) and CNNs for accurate perturbation removal while preserving background contents. Based on this, we carry forward a Retinex-based framework, dubbed as Mutual Retinex, where a two-branch structure is designed to characterize the specific knowledge of reflectance and illumination components while removing the perturbation. To maximize its potential, Mutual Retinex is equipped with a new mutual learning mechanism, involving an elaborately designed mutual representation module (MRM). In MRM, the complementary information between reflectance and illumination components are encoded and used to refine each other. Through the complementary learning via the mutual representation, the enhanced results generated by our model exhibit superior color consistency and naturalness. Extensive experiments have shown the significant superiority of our mutual learning based method over thirteen competitors on the low-light task and ten methods on the underwater image enhancement task. In particular, our proposed Mutual Retinex respectively surpasses the state-of-the-art method MIRNet-v2 by 0.90 dB and 2.46 dB in PSNR on the LOL 1000 and FIVEK datasets, while with only 19.8% model parameters. © 2017 IEEE.

关键词： Image enhancement

来源：评论

学校读者我要写书评

暂无评论

Beyond Preferences: Enriching User Profiles for Effective E-commerce Recommendations

Beyond Preferences: Enriching User Profiles for Effective E-...

引用

IEEE International Symposium on Circuits and Systems (ISCAS)

作者： Jingyu Xu Zhengwei Yang Zheng Wang National Engineering Research Center for Multimedia Software School of Computer Science Wuhan University

ISBN: (数字)9798350356830

ISBN: (纸本)9798350356847

Recommender systems are essential in E-Commerce platforms, with recent advancements leveraging users’ historical records to extract multi-interests. However, beyond these records, user profiles contain semantic information that inherently shapes their interests. Existing works mainly overlook that a user’s interests have: group influence, multi-level preference, and time relevance. To this end, a novel Enhanced User Profile-based Multi-interest Model (E-UPMiM) for recommendation is proposed to integrate enhanced user profiles with social relationships to model users’ multi-interests effectively. We propose to extract user preferences with three components: 1) integrated input containing enhanced profiles with social relationships to meet users’ grouping needs; 2) a multi-interest extraction module to obtain complex interest representations; and 3) a time-aware ranking module to adjust the recommendations dynamically. Extensive experiments on three public datasets show that E-UPMiM significantly outperforms state-of-the-art recommendation models. Codes are publicly available at: https://***/KevinXu-01/E-UPMiM.

关键词： Codes Accuracy Shape Graph convolutional networks Circuits and systems Semantics Electronic commerce User preference Integrated circuit modeling Recommender systems

来源：评论

学校读者我要写书评

暂无评论

Sensor-Based Authentication on Smartphones via Integrating Auxiliary Information

Sensor-Based Authentication on Smartphones via Integrating A...

引用

2024 IEEE International Conference on Systems, Man, and Cybernetics, SMC 2024

作者： Wang, Yingjie Hu, Ruimin Wuhan University National Engineering Research Center for Multimedia Software Whuhan430072 China School of Cyber Science and Engineering Wuhan University Key Laboratory of Aerospace Information Security and Trusted Computing Ministry of Education Whuhan430072 China

ISBN: (纸本)9781665410205

With the widespread use of smartphones, more and more private information is stored on the phone, and the loss or theft of the phone can lead to data leakage, theft of property, and other problems. Traditional active authentication methods based on passwords, faces, fingerprints, etc. authenticate only once at login, which still has some hidden dangers. Authentication methods based on behavioral biometrics such as walking gait, touch screen, keystroke, etc. can provide implicit and continuous authentication services, and thus have been widely studied recently. Considering the limited computational resources of smart devices, we first introduce the state-space model with linear model complexity, i.e., Mamba, to extract deep feature representations from raw signal sequences in the sensor-based authentication task. Then a series of auxiliary information from different domains are proposed and fused to the deep features to enhance the consistent semantic information and ultimately improve the discriminative ability of the model. Our model A-Mamba, i.e., Auxiliary-Mamba, is experimented on two public datasets. The experimental results show that our model outperforms existing approaches. © 2024 IEEE.

关键词： Information leakage

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：