检索结果-内蒙古大学图书馆

27th International Conference on Pattern Recognition, ICPR 2024

作者： Lin, Xiaowen Aysa, Alimjan Ubul, Kurban School of Computer Science and Technology Xinjiang University Ürümqi China Joint International Research Laboratory of Silk Road Multilingual Cognitive Computing Ürümqi China Xinjiang Key Laboratory of Multilingual Information Technology Ürümqi China

ISBN: (纸本)9783031781681

In recent years, prompt-based learning has achieved some success in various natural language tasks. In text classification tasks, the construction of prompt templates and label mapping have a significant impact on the results. Therefore, many methods for constructing label mappings have emerged, such as manual and automatic construction. However, manual label construction is time-consuming, expensive, and has low scalability, while automatic label construction requires a large number of samples and other limitations. Therefore, we have explored a automatic label mapping method that only requires one sample. On the basis of only requiring one sample, we achieved good experimental results. In this paper, we propose an effective one-shot method for automatically constructing label mapping, ALaM. Our method is based on statistical algorithms and label semantic information for one to many label mapping. Our experiments demonstrate that ALaM achieves competitive performance on the GLUE benchmark without human effort or external resources. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

关键词： Zero-shot learning

来源：评论

学校读者我要写书评

暂无评论

Oracle Character Recognition Based on Attention Enhancement and Multi-level Feature Fusion 27th

Oracle Character Recognition Based on Attention Enhancement...

引用

27th International Conference on Pattern Recognition, ICPR 2024

作者： Han, Zhiwang Yadikar, Nurbiya Xu, Xuebin Aysa, Alimjan Ubul, Kurban School of Computer Science and Technology Xinjiang University Urumqi830046 China Xinjiang Multilingual Information Technology Key Laboratory Xinjiang University Urumqi830046 China Joint International Research Laboratory of Silk Road Multilingual Cognitive Computing Xinjiang University Urumqi830046 China

ISBN: (纸本)9783031781186

Oracle bone characters represent the earliest inscriptions in China. Recognizing and deciphering these characters is significant. Despite some progress made by recent methods, their recognition accuracy remains limited by two major issues: 1) how to focus on characters features within complex background noise images, and 2) how to effectively fuse shallow detail information with deep semantic information. To address these issues, we propose a novel deep learning model called Character Feature Enhancement Network (CFE-Net). The model consists of two key components: Character Feature Enhancement (CFE) and Adaptive Multi-level Classifier Fusion (AMCF). Specifically, CFE utilizes the Spatial Focus Attention Module (SFAM) to focus on extracting foreground character features and suppressing background noise, thereby significantly enhancing high-level semantic representation capabilities. AMCF, on the other hand, achieves multi-level feature fusion by adaptively fusing the outputs of different classifiers, effectively avoiding information loss or interference that simple fusion strategies might cause. We evaluated the CFE-Net on two rubbing oracle bone characters benchmark datasets, OBC306 and Oracle-MNIST. The experimental results demonstrate that CFE-Net significantly outperforms several existing methods in terms of Top-1 accuracy, establishing it as the new state-of-the-art. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

关键词： Semantics

来源：评论

学校读者我要写书评

暂无评论

Oracle Bone Inscription Image Retrieval Based on Improved ResNet Network 27th

Oracle Bone Inscription Image Retrieval Based on Improved Re...

引用

27th International Conference on Pattern Recognition, ICPR 2024

作者： Ding, Jun Wang, Jiaoyan Aysa, Alimjan Xu, Xuebin Ubul, Kurban School of Computer Science and Technology Xinjiang University Ürümqi830046 China Xinjiang Multilingual Information Technology Key Laboratory Xinjiang University Ürümqi830046 China Joint International Research Laboratory of Silk Road Multilingual Cognitive Computing Xinjiang University Urumqi830046 China

ISBN: (纸本)9783031783043

Research on oracle bone inscription image retrieval is important for applications in academic and cultural heritage areas. The current oracle bone dataset faces problems such as the low similarity between the same category, the high similarity between the different categories, and imbalanced sample distribution. In addition, due to the complex background of oracle bone images, existing network models have certain limitations in extracting image features. To address these challenges, this study first adopts a Siamese network-based image retrieval method to learn feature representations of similar and dissimilar images. Subsequently, the existing dataset was partitioned, providing a practical and usable retrieval dataset for the oracle bone image retrieval field. Finally, an improved network model based on ResNet is proposed and integrated into the Siamese network framework. The model achieves the highest retrieval MP and MAP values of 83.26% and 90.68%, respectively, which is better than the current research. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

关键词： Content based retrieval

来源：评论

学校读者我要写书评

暂无评论

Oracle Bone Script Recognition Based on Multi-scale Feature Fusion and Knowledge Distillation 27th

Oracle Bone Script Recognition Based on Multi-scale Feature ...

引用

27th International Conference on Pattern Recognition, ICPR 2024

作者： Wang, Jiaoyan Xu, Xuebin Aysa, Alimjan Ding, Jun Ubul, Kurban School of Computer Science and Technology Xinjiang University Ürümqi830046 China Xinjiang Multilingual Information Technology Key Laboratory Xinjiang University Ürümqi830046 China Joint International Research Laboratory of Silk Road Multilingual Cognitive Computing Xinjiang University Ürümqi830046 China

ISBN: (纸本)9783031783111

Oracle bone scripts, being the oldest and most developed writing system discovered in China till date, have been meticulously studied by numerous scholars. This paper presents the Multi–scale Feature Fusion Attention Net (MFFA–Net) as a solution to the issue of variant characters in the oracle bone affecting recognition accuracy. The network is based on ResNet18 and employs asymmetric convolution alongside a refined coordinate attention technique to acquire image features. It also integrates perceptual field information of varying sizes through hierarchical bilinear pooling. Furthermore, knowledge from the Wide_ResNet101 and DenseNet169 is transferred to MFFA-Net using knowledge distillation techniques. Finally, to validate the effectiveness of our proposed method, we conducted rigorous experiments on the OBC306, OBC265, and EOBC datasets, comparing our results with those of existing methods. The experimental results demonstrate that our method obtains remarkable performance in oracle character recognition, reaching state-of-the-art Top–1 accuracy with 92.42%, 94.78%, and 98.82% on the OBC306, OBC265, and EOBC datasets, respectively. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

关键词： Attention Feature Fusion Knowledge Distillation Oracle Recognition

来源：评论

学校读者我要写书评

暂无评论

Improved Cross-Lingual Speaker Verification Using Speaker Sensitive Feature Guidance and Fine-grained Phonetic Information

Improved Cross-Lingual Speaker Verification Using Speaker Se...

引用

International Conference on Acoustics, Speech, and Signal Processing (ICASSP)

作者： Yongtai Ji Guangxing Li Hao Huang Yanbing Li Wushour Silamu School of Computer Science and Technology Xinjiang University Urumqi China Joint International Research Laboratory of Silk Road Multilingual Cognitive Computing Urumqi China Xinjiang Key Laboratory of Multi-Lingual Information Technology Urumqi China

ISBN: (数字)9798350368741

ISBN: (纸本)9798350368758

Speaker verification performance significantly degrades when there exists a language mismatch between training and evaluation. Domain Adversarial Training (DAT) has shown to be effective in mitigating this gap by incorporating adversarial training with domain information (language id). Inspired by recent research of DAT, we propose to decompose and locate features sensitive to speaker identity, so that domain adaptation can better serve the speaker classification task. Additionally, we further promote DAT by replacing alignment based on language identification with alignment of fine-grained phonetic information, where pre-trained speech recognition models are utilized to provide frame-level phonetic labels. The speaker verification model is trained on VoxCeleb, while CnCeleb is used for adversarial training and evaluation. Results show the methods effectively mitigate the performance degradation caused by language mismatch.

关键词： Training Degradation Adaptation models Visualization Speech recognition Phonetics Signal processing Acoustics Speech processing

来源：评论

学校读者我要写书评

暂无评论

HANet: A Harmonic Attention-Based Network for Singing Melody Extraction from Polyphonic Music

HANet: A Harmonic Attention-Based Network for Singing Melody...

引用

International Conference on Acoustics, Speech, and Signal Processing (ICASSP)

作者： Shijun Wang Xiangzhu Kong Hao Huang Kai Wang Ying Hu School of Computer Science and Technology Xinjiang University Urumqi China Joint International Research Laboratory of Silk Road Multilingual Cognitive Computing Urumqi China Xinjiang Key Laboratory of Multi-Lingual Information Technology Urumqi China

ISBN: (数字)9798350368741

ISBN: (纸本)9798350368758

Singing melody extraction from polyphonic music is a complex but important task in music information retrieval. Harmonic relationships have been shown to be crucial in this task, but most existing models based on Convolutional Neural Networks (CNNs) struggle to capture long-range harmonic dependencies. To address this, we propose a Harmonic Attention-based Network (HANet) for singing melody extraction from polyphonic music, which includes multiple sampling layers. Specifically, each sampling layer uses three parallel Harmonic Attention Modules (HAMs) with CNNs of different kernel sizes to capture harmonic relationships across 1 to 6 octaves along the frequency axis. Additionally, the Channel Attention Module (CAM) is used to adaptively model long-range harmonic features across various octave ranges. Experimental results show that HANet achieves state-of-the-art performance on multiple public music melody datasets. The overall accuracy on the ADC2004 dataset reached a peak of 95.2%, with the lowest voicing false alarm result of 2.1%. Visualization results directly demonstrate that harmonic information effectively reduces octave errors. Our code is available online: https://***/wwsjj/HANet

关键词： Measurement Visualization Accuracy Signal processing Harmonic analysis Convolutional neural networks Data mining Kernel Speech processing Music information retrieval

来源：评论

学校读者我要写书评

暂无评论

Robust and Efficient Text-based Speech Editing using Noise Conditioning and Rectified Flow

Robust and Efficient Text-based Speech Editing using Noise C...

引用

International Conference on Acoustics, Speech, and Signal Processing (ICASSP)

作者： Haowen Yin Kai Wang Hongli Yang Hao Huang Wushour Silamu School of Computer Science and Technology Xinjiang University Urumqi China Xinjiang Key Laboratory of Multi-Lingual Information Technology Urumqi China Joint International Research Laboratory of Silk Road Multilingual Cognitive Computing Urumqi China

ISBN: (数字)9798350368741

ISBN: (纸本)9798350368758

Significant advancements have been made in text-based speech editing (TSE) for clear speech, but effectively editing the noise-contaminated speech remains a challenge. Background noise degrades the quality of generated speech, and edited speech that fails to maintain noise context consistency often sounds unnatural. We propose Reflow-TSE, a robust and efficient TSE model for noise-consistent speech editing. For noise robustness, 1) we design a noise condition module to extract frame-level noise sequences as the conditioning information, and 2) we introduce an enhanced context-conditioned prediction module that predicts masked noise sequences along with conventional duration and pitch, using these conditions to guide generation. For boosting efficiency, 3) we introduce the rectified flow model, leveraging speech context and predicted conditions to achieve high-quality editing with limited sampling steps. Experimental results show that with just two steps of sampling, Reflow-TSE achieves context-consistent noisy speech editing, a capability absent in other TSE models. Additionally, for clean speech, Reflow-TSE also matches or surpasses baseline models. Audio samples are available at https://***.

关键词： Noise Predictive models Boosting Acoustics Noise robustness Noise measurement Human voice Data mining Background noise Context modeling

来源：评论

学校读者我要写书评

暂无评论

Utterance as A Bridge: Few-shot joint Learning of Empathy Detection and Empathy Intent Classification

Utterance as A Bridge: Few-shot Joint Learning of Empathy De...

引用

International Conference on Acoustics, Speech, and Signal Processing (ICASSP)

作者： Liting Jiang Di Wu Zhe Li Shuangyong Song Yanbing Li Hao Huang School of Computer Science and Technology Xinjiang University Urumqi China Xinjiang Key Laboratory of Multi-Lingual Information Technology Urumqi China Institute of Artificial Intelligence (TeleAI) China Telecom Corp Ltd Beijing China Department of Electrical and Electronic Engineering Hong Kong Polytechnic University Hong Kong China Department of Electrical Engineering Stanford University USA Joint International Research Laboratory of Silk Road Multilingual Cognitive Computing Urumqi China

ISBN: (数字)9798350368741

ISBN: (纸本)9798350368758

Empathy detection (ED) and empathy intent classification (EIC) aim to identify the empathy direction expressed in user utterances and the underlying empathy intent behind them. Previous studies show that facilitating information transfer between tasks can enhance model performance. However, the interaction between ED and EIC in few-shot learning remains underexplored. To this end, we identify the challenges in jointly training ED and EIC in a few-shot setting: establishing effective information transfer between them and improving the model’s generalization capability. We propose a novel model called USB. For information transfer, the interactive module maps empathy and empathy intent labels through utterances to model task correlations. For generalization capability, after capturing empathy and empathy intent representations with an adaptive fusion module, we introduce a multi-level contrastive learning strategy to optimize representations at task and label levels, enhancing generalization. Experimental results on two public datasets show that our model outperforms all baselines.

关键词： Training Bridges Adaptation models Intent recognition Focusing Contrastive learning Signal processing Universal Serial Bus Few shot learning Speech processing

来源：评论

学校读者我要写书评

暂无评论

MDH-NAS: Accelerating EEG Signal Classification with Mixed-Level Differentiable and Hardware-Aware Neural Architecture Search

引用

IEEE Internet of Things Journal 2025年

作者： Zhu, Lixian Wang, Su Jin, Xiaokun Zheng, Kai Zhang, Jian Sun, Shuting Tian, Fuze Cai, Ran Hu, Bin Key Laboratory of Brain Health Intelligent Evaluation and Intervention Beijing Institute of Technology Beijing100081 China Beijing Institute of Technology School of Medical Technology Beijing100081 China Lanzhou University School of Information Science and Engineering Lanzhou730000 China Lanzhou University Gansu Provincial Key Laboratory of Wearable Computing School of Information Science and Engineering Lanzhou China Chinese Academy of Sciences CAS Center for Excellence in Brain Science and Intelligence Technology Shanghai Institutes for Biological Sciences Shanghai China Chinese Academy of Sciences Joint Research Center for Cognitive Neurosensor Technology of Lanzhou University Institute of Semiconductors Lanzhou China

In non-invasive brain-computer interfaces (BCIs), EEG analysis plays a critical role, with neural networks serving as a cornerstone for signal decoding. Existing neural network approaches for EEG signal recognition require extensive manual design and hyperparameter tuning, leading to inefficiencies and making them impractical for embedded devices due to their large model size. To address these limitations, we propose Mixed-Level Differentiable and Hardware-Aware Neural Architecture Search (MDH-NAS), a framework that automatically generates lightweight neural networks tailored for EEG classification. Unlike traditional DARTS methods, MDH-NAS employs a hybrid optimization strategy that balances global and local search spaces, thereby accelerating and refining architecture discovery. It introduces explicit size constraints during the search process to ensure deployability on embedded devices. MDH-NAS demonstrates autonomous generation of architectures for tasks such as motor imagery (MI) and depression recognition, achieving 87.80% accuracy on the BCI-IV dataset and 90.09% on the MODMA dataset. When deployed on the EAIDK-610 board across heterogeneous tasks, it attains 85.37% accuracy on the EEG Motor Movement/Imagery dataset. This method reduces architecture discovery time by 89% and enhances prediction accuracy by 8.70% compared to baseline methods, highlighting its potential for scalable EEG analysis and real-world embedded deployment. © 2014 IEEE.

关键词： Electroencephalography

来源：评论

学校读者我要写书评

暂无评论

AugReID: Transformer-Based Augmentation Person Re-Identification 7

AugReID: Transformer-Based Augmentation Person Re-Identifica...

引用

7th Asian Conference on Artificial Intelligence technology, ACAIT 2023

作者： Yu, Zaiyang Zhang, Liping Sun, Linjun Ning, Xin Li, Weijun Institute of Semiconductors Chinese Academy of Sciences Beijing China University of Chinese Academy of Science China Cognitive Computing Technology Joint Laboratory Wave Group Beijing China

ISBN: (纸本)9798350359145

Person re-identification (ReID) is a crucial component of computer vision that has garnered increased attention in recent years, primarily due to its importance in applications such as video surveillance. However, existing research predominantly focuses on ReID performance in ideal dataset environments, often overlooking the challenges associated with implementing ReID algorithms in real-world deployment scenarios. In these real-world situations, variations in light intensity and resolution between edge-end cameras can considerably hinder the accurate interpretation of person characteristics. To address these challenges, we propose an augmentation strategy to simulate interference in real-world deployment scenarios. Furthermore, we introduce a noise fusion model structure, AugReID, designed to account for the impact of illumination, resolution, and other factors on the ReID task. Extensive experiments on holistic ReID datasets demonstrate our model's effectiveness. Additionally, our approach surpasses state-of-the-art methods on the Occluded ReID task, which contains more noise-related elements and closely resembles real deployment environments. © 2023 IEEE.

关键词： Computer vision

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：