检索结果-内蒙古大学图书馆

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

作者： Yang Xiao Liejun Wang Tongguan Wang Huicheng Lai Xinjiang Key Laboratory of Signal Detection and Processing College of Information Science and Engineering Xinjiang University Urumqi China

Violence detection is an application of anomaly detection, which is used to detect violence content in video clips. Using multimodal as input can improve the performance of violence detection. However, the existing MML Transformers-based fusion methods do not take into account the differences between non-homologous modals. The fusion of non-homologous modals makes features become noise between each other. This paper proposes a score fusion-based transformer framework, named Scoreformer. First of all, the optical flow, RGB and audio features pass through the independent self-attention transformer blocks. Second, the optical flow and RGB features pass through the cross-modal transformer blocks, after that they are fused with the audio features through the score fusion block. This method avoids the noise interference caused by the direct fusion of audio features and visual features. Experiments on the XD-Violence dataset show that the proposed method achieves 84.54% of the AP value, which exceeds at least 2.85% compared with the most advanced method (e. g. MSL, CRFD).

关键词： Visualization Interference signal processing Transformers Feature extraction Mobile handsets Explosions

来源：评论

学校读者我要写书评

暂无评论

Low-Resource Speech keyword Search Based on Residual Neural Network 16th

Low-Resource Speech Keyword Search Based on Residual Neural ...

引用

16th Chinese Conference on Biometric Recognition, CCBR 2022

作者： Wang, Dafei Huang, Zhihua Li, Hui Liu, Wenchen School of Information Science and Engineering Xinjiang University Urumqi830017 China Key Laboratory of Signal Detection and Processing in Xinjiang Uygur Autonomous Region Xinjiang University Urumqi830017 China

ISBN: (纸本)9783031202322

Low-resource speech keyword search is usually based on DTW template matching. However, this method has low precision and is sensitive to speaker or recording environment changes. This paper proposes a low-resource speech keyword search method based on image recognition. Bottleneck features are extract through a pre-trained model based on multi-language to generate similarity matrix images. ResNet is used to classify similarity matrix images. According to the image classification result, it is determined whether the speech keyword appears. This model does not require prior language knowledge and abundant corpus resources, so it is suitable for low-resource speech keyword search. We validate the performance of the model on different datasets. On SWS 2013, the model recall, accuracy, and F1 increased by 20.83%, 17.59%, and 19.18%, respectively. On QUESST 2014, they increased by 24.44%, 24.96%, and 24.70%, respectively. Experimental results show that the propose method is effective. © 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.

关键词： Template matching

来源：评论

学校读者我要写书评

暂无评论

Pipeline Chain-of-Thought: A Prompt Method for Large Language Model Relation Extraction

Pipeline Chain-of-Thought: A Prompt Method for Large Languag...

引用

International Conference on Asian Language Processing (IALP)

作者： Hangtian Zhao Hakiz Yilahun Askar Hamdulla Xinjiang Key Laboratory of Signal Detection and Processing School of Computer Science and Technology XinJiang University Urumqi China

The development of language models has been influencing approaches to relation extraction (RE) problems. Although large language models (LLMs) have demonstrated breakthrough potential in certain aspects, they are still in the exploratory stage for RE tasks. Currently, the mainstream approach to improving the RE performance of LLMs is through prompt fine-tuning, but most methods require providing entity information in the prompt, which effectively only allows the LLMs to perform relationship classification tasks. We propose the Pipeline Chain-of-Thought (Pipeline-COT), which breaks down the RE task into steps and transforms it into reasoning tasks that have flat scaling curves, thereby enabling the use of Chain-of-Thought (COT) to enhance model inference. In addition, our method utilizes n-shot samples to provide signals for the Bayesian inference of the model by prompting the LLMs to focus on specific concepts to generate answers. We evaluated pipeline-COT on the Chinese dataset DuIE2.0, and compared with baseline methods that require including entity information in the prompt, our method still shows competitive performance.

关键词：

来源：评论

学校读者我要写书评

暂无评论

High-quality endoscopic laser speckle contrast imaging with conical fiber illumination

引用

Photonics Research 2025年第3期13卷 583-592页

作者： JUNSHUAI YAN QINXIN HAN LIANGWEI MENG TINGYU SUN YAN YAN SHIJIE FENG SHAOMIN YUAN JINLING LU PENGCHENG LI Britton Chance Center for Biomedical Photonics-MoE Key Laboratory for Biomedical Photonics Advanced Biomedical Imaging Facility-Wuhan National Laboratory for OptoelectronicsHuazhong University of Science and TechnologyWuhan 430074China Research Unit of Multimodal Cross Scale Neural Signal Detection and Imaging Chinese Academy of Medical ScienceHUST-Suzhou Institute for BrainsmaticsJITRISuzhou 215100China State Key Laboratory of Digital Medical Engineering Key Laboratory of Biomedical Engineering of Hainan ProvinceSchool of Biomedical EngineeringHainan UniversitySanya 572025China

Blood flow is essential for maintaining normal physiological functions of the human *** laser speckle contrast imaging(LSCI)can achieve rapid,high-resolution,label-free,and long-term blood flow perfusion velocity monitoring in minimally invasive ***,conventional endoscopic LSCI uses a lowcoherence laser illumination scheme,leading to restricted angles of illumination,compromised laser coherence,uneven laser illumination distribution,and low coupling efficiency,all of which degrade the quality of LSCI in the *** this paper,we propose that conical fiber(CF)-coupled high-coherence laser can be used to achieve large-angle,high-coherence,high-uniformity,and high coupling efficiency laser illumination in the ***,we establish an effective model for calculating the divergence angle of *** phantom and animal experiments,we reveal that laser illumination based on CF markedly enhances endoscopic LSCI *** technology broadens the imaging field of view,enhances the signal-to-noise ratio,enables more sensitive detection of minute blood flow changes,expands the detectable flow range,and improves signal-to-background ratio of endoscopic *** findings suggest that CF-based laser illumination stands as a highly promising advancement in endoscopic LSCI.

关键词： conical fiber maintaining normal physiological functions lowcoherence laser illumination schemeleading laser speckle contrast imaging lsci can endoscopic laser speckle contrast imaging blood flow blood flow perfusion velocity monitoring high coherence laser

来源：评论

学校读者我要写书评

暂无评论

Research of Pitch Controllability Based on LPCNet 3

Research of Pitch Controllability Based on LPCNet

引用

3rd International Conference on Pattern Recognition and Machine Learning, PRML 2022

作者： Gao, Wenyu Hamdulla, Askar Yang, Xipeng Chen, Yunlin School of Information Science and Engineering Xinjiang University of China China Xinjiang Key Laboratory of Signal Detection and Processing Xinjiang Urumqi China Xinjiang Key Laboratory of Multilingual Information Processing China Mobvoi Ai Lab Suzhou China

ISBN: (数字)9781665499507

ISBN: (纸本)9781665499507

With the development of speech synthesis technology in most scenarios, modifying the pitch of audio has been widely used. Currently most pitch controllable methods are achieved by using a separate reference encoder neural network, but this approach requires training more complicated neural networks and does not generalize to unseen speakers without speaker adaptation. To address this problem, a method based on LPCNet vocoder for explicitly pitch control is proposed. Firstly, during the training phase, by optimizing the training features, the model can synthesize more natural and robust voices. Next, the fundamental frequency is adjusted by inputting a control curve in the inference stage, and then the acoustic features are inferred into waveform using the vocoder LPCNet. The experimental results show that the proposed method can control the pitch flexibly and the sound quality of the synthetic voice is improved. © 2022 IEEE.

关键词： Speech synthesis

来源：评论

学校读者我要写书评

暂无评论

Fake News detection Based on Cross-modal Co-attention

Fake News Detection Based on Cross-modal Co-attention

引用

Pattern Recognition and Machine Learning (PRML), IEEE International Conference on

作者： Yadong Gu Mijit Ablimit Askar Hamdulla School of Information Science and Engineering Xinjiang University Xinjiang Key Laboratory of Signal Detection and Processing Urumqi China

The spread of rumors has brought many negative impacts to society. Nowadays, rumors on social media platforms often exist in the form of both images and text. In response to this, many methods for multimodal rumor detection have been proposed, but most of these methods only utilize extracted text features and image features for simple fusion, without considering the connections between modalities. In this study, we employ ConvNet to extract image features, BERT to extract text features, and take into account the complementarity between modalities. We use a co-attention module to enhance the extracted features of individual modalities, aiming to achieve the interaction of modality features and the complementary nature between modality information. The obtained features are then concatenated to achieve the fusion of image and text features. The multimodal fusion model used in this study was experimented on Twitter and Weibo, achieving accuracy rates of 81.20% and 88.15% respectively.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Whole-Brain Connectome of GABAergic Neurons in the Mouse Zona Incerta

引用

Neuroscience Bulletin 2022年第11期38卷 1315-1329页

作者： Yang Yang Tao Jiang Xueyan Jia Jing Yuan Xiangning Li Hui Gong Britton Chance Center for Biomedical Photonics Wuhan National Laboratory for OptoelectronicsMoE Key Laboratory for Biomedical PhotonicsHuazhong University of Science and TechnologyWuhan 430074China Research Unit of Multimodal Cross Scale Neural Signal Detection and Imaging Chinese Academy of Medical SciencesHUST-Suzhou Institute for BrainsmaticsJITRISuzhou 215123China

The zona incerta(ZI)is involved in various functions and may serve as an integrative node of the circuits for global behavioral ***,the long-range connectivity of different sectors in the mouse ZI has not been comprehensively ***,we obtained whole-brain images of the input and output connections via fluorescence micro-optical sectioning tomography and viral *** principal regions in the input-output circuits of ZI GABAergic neurons were topologically *** 3D distribution of cortical inputs showed rostro-caudal correspondence with different ZI sectors,while the projection fibers from ZI sectors were longitudinally organized in the superior *** results show that the medial and lateral ZI are two different major functional compartments,and they can be further divided into more subdomains based on projection and input *** study provides a comprehensive anatomical foundation for understanding how the ZI is involved in integrating different information,conveying motivational states,and modulating global behaviors.

关键词： Zona incerta GABAergic neurons Whole-brain connectome Input circuit Output circuit Topological connection

来源：评论

学校读者我要写书评

暂无评论

Efficient Channel Attention for Speech Emotion Recognition With Asymmetric Convolution Module

Efficient Channel Attention for Speech Emotion Recognition W...

引用

Pattern Recognition and Machine Learning (PRML), IEEE International Conference on

作者： TianQi Wu Liejun Wang Shuli Cheng Jiang Zhang School of Information Science and Engineering Xinjiang University Xinjiang Key Laboratory of Signal Detection and Processing Urumqi China

Speech emotion recognition (SER) is a challenging task due to the diversity and complexity of emotions. There are some unresolved issues, such as too much redundant information in the extracted features, the problems of hard-to-classify samples and category imbalance. To address these issues, we propose asymmetric convolution module (ACM) capture the spatial information of feature maps by improving CNNs structure and efficient channel attention (ECA) to establish the dependency relationship between channels. The combination of both extract more useful emotion information in the feature map and improve classification accuracy. In this paper, our proposed model achieves state-of-the-art results on IEMOCAP improvised data, with 70.99% and 89.78% unweighted accuracy (UA) on the full IEMOCAP and EMO-DB corpora.

关键词：

来源：评论

学校读者我要写书评

暂无评论

A Standardized-Based Knowledge Distillation Model for Multimodal Emotion Recognition in Conversation

SSRN

引用

SSRN 2024年

作者： Li, Yihao Tohti, Turdi Han, Dongfang Liang, Yi Zuo, Zicheng Liao, Yuanyuan Yang, Qingwen School of Computer Science and Technology Xinjiang University Urumqi830017 China Xinjiang Key Laboratory of Signal Detection and Processing Urumqi830017 China

In recent years, with the rapid development of artificial intelligence and deep learning technology, Multimodal Emotion Recognition in Conversation (MERC) has gradually attracted widespread attention. However, existing methods rely on designed models to learn modal representations, there is a lack of research on enhancing modal representations. Therefore, we propose a method for standardized knowledge distillation. This method uses the intra-modal Transformer module for learning intra-modal representations, as well as employing the Cross-Modal Transformer module to effectively integrate different modalities. In addition, we further design a novel standardized knowledge distillation strategy, which uses the soft labels of the text modality as training supervision, thereby achieving the migration of semantic and contextual information from the text modality to the audio and visual modalities. This strategy fully borrows the idea of information sharing, allowing the audio and visual modalities to learn the fine-grained features and complex decision boundaries of the text modality. To demonstrate the effectiveness of our proposed model, we conduct extensive experiments on two public emotion recognition datasets: IEMOCAP and MELD. Experimental results further demonstrate the superiority of our method compared with the state-of-the-art methods. © 2024, The Authors. All rights reserved.

关键词： Distillation

来源：评论

学校读者我要写书评

暂无评论

A Physical Information Neural Network Algorithm for Centroid Solution of Cross Strip Anode Detector 4

A Physical Information Neural Network Algorithm for Centroid...

引用

4th International Conference on Electronic Information Engineering and Computer Communication, EIECC 2024

作者： Zuo, Xiaoyun Zheng, Jinkun Bai, YongLin Duan, Jinyao Miao, Yixin Qiao, Cong Key Laboratory of Ultra-fast Photoelectric Diagnostics Technology of Chinese Academy of Sciences Xi'an Institute of Optics and Precision Mechanics Chinese Academy of Sciences Xi'an China The University of Chinese Academy of Sciences Beijing China The Laboratory of Space Science Weak-signal Detection Technology Key Laboratory of Ultra-fast of Chinese Academy of Sciences Xi'an Institute of Optics and Precision Mechanics Chinese Academy of Sciences Xi'an China Xi'an University of Posts and Telecommunications Xi'an China

ISBN: (纸本)9798331534622

With the rapid development of aerospace technology, the detection ability of remote dim target has become a key task. Cross strip anode (XS) photoelectric imaging detector can be used to capture weak target signal, has become a powerful tool for remote target detection. Due to the complex and changeable space environment, various noises will inevitably enter the detection system, thus interfering with the accurate detection of targets and reducing signal resolution. In order to improve the precision of calculating the centroid of cross strip anode under strong noise, a model based on physical information neural network (PINN) is proposed. Compared with the traditional centroid method, the physical information based neural network method does not need to set specific parameters according to the type of noise, so it is hardly affected by noise interference and can obtain high precision centroid localization. © 2024 IEEE.

关键词： Anodes

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：