检索结果-内蒙古大学图书馆

Mixformer: Feature Mixed Transformer for Rainfall Forecasting

学校读者我要写书评

暂无评论

Mixformer: Feature Mixed Transformer for Rainfall Forecastin...

IEEE International Conference on Systems, Man and Cybernetics

作者： Yuanyuan Liao Boyuan Li Xiuhong Li School of Computer Science and Technology Xinjiang University Urumgi China Key Labaratory of Signal Detection and Processing. Xinjiang Uygur Auonomous Region Xinjiang University Urumqi China

ISBN: (数字)9781665410205

ISBN: (纸本)9781665410212

In the xinjiang region of China, water is scarce and unevenly distributed, and due to factors such as global warming, extreme rainfall events occur frequently, posing serious threats to people's lives and property safety. To accurately predict short-term precipitation, we propose a Mixformer model. Specifically, we first use a combination of min-max normalization and reversible instance normalization to reduce the impact of feature numerical ranges on modeling while preserving the distribution of the original data. Next, to enhance the representation of multivariate time series data, we propose a feature mixing module. This module enhances the representation of inter-feature correlation information by calculating the correlation between different sequences, thus improving prediction effectiveness. Finally, to enhance its non-linear characteristics, we introduce a residual prediction method. This method models seasonal and trend components separately and also pays special attention to the residual component. We have validated the proposed method extensively, proving that it outperforms existing state-of-the-art (SOTA) methods in this field.

关键词： Water Correlation Rain Accuracy Time series analysis Predictive models Transformers Market research Numerical models Safety

Boundary-Guided Fusion of Multi-Level Features Network for Camouflaged Object detection

学校读者我要写书评

暂无评论

Boundary-Guided Fusion of Multi-Level Features Network for C...

International Joint Conference on Neural Networks (IJCNN)

作者： Songlin Li Zhe Li Boyuan Li Xiuhong Li Jiabao Sheng School of Computer Science and Technology Xinjiang University Urumqi China Xinjiang Key Laboratory of Signal Detection and Processing Xinjiang University Urumqi China Department of Electrical and Electronic Engineering Hong Kong Polytechnic University HongKong China

ISBN: (数字)9798350359312

ISBN: (纸本)9798350359329

Camouflaged objects, exhibiting high similarity with their surroundings, pose a substantial challenge for both humans and machines to detect when concealed within the environment. Existing methods for camouflage object detection (COD) struggle in accurately segmenting the overall structure of camouflaged objects. To address this issue, we propose a novel boundary-guided fusion of multi-level features network (BGFM-Net) for COD. In contrast to existing boundary-guided methods, we pay more attention to addressing the significant imbalance in the pixel quantities between boundary and background features, allowing for a more comprehensive representation of boundary features. BGFM-Net primarily consists of a multi-scale aggregation module (MSAM), a boundary-guided feature module (BFM), and a cross-Level fusion module (CLFM). MSAM effectively integrates contextual semantics at different scales, achieving a powerful and efficient feature representation. BFM adeptly combines edge features while constraining interference from background features, guiding the learning of camouflaged object boundary representation. CLFM integrates multi-level features for predicting camouflaged objects while adaptively adjusting channel weights to emphasize important channels and diminish the impact of less relevant channels for the task. Extensive experiments on three benchmark camouflage datasets demonstrate that our BGFM-Net outperforms other state-of-the-art COD models.

关键词： Adaptation models Semantics Neural networks Object detection Interference Benchmark testing Feature extraction

Joint Colour Casts Correction and Zero-Shot Learning for Real-World Sand-Dust Image Enhancement

学校读者我要写书评

暂无评论

SSRN 2024年

作者： Shi, Fei Ren, Dayong Jia, Zhenhong Zhou, Yanyun School of Information Science and Engineering Xinjiang University Urumqi830046 China Key Laboratory of Signal Detection and Processing Xinjiang University Urumqi830046 China National Key Laboratory for Novel Software Technology Nanjing University Nanjing210023 China

Low illumination and various colour casts are present in images taken in sand-dust weather. A deep learning-based or prior-based desanding approach may not produce the best results. This work proposes an efficient joint colour cast correction and zero-shot learning framework to formulate the image desanding problem as the task of improving image visibility and contrast. In the colour cast correction stage, by using the compensation between each colour channel, Based on the statistical properties of sand dust images, an adaptive colour cast correction is carried out. At the contrast enhancement stage, we present an end-to-end illumination map estimate method based on zero-shot learning. First, we propose a novel method for generating a degraded image by controlling the perturbation of the Retinex model. Then, the input and degraded images are used to train the model, which is then used to make the luminance map. Finally, combining all configuration constraints gives a good visual quality enhancement. Evaluations on a large number of different sand dust images, especially those with significant colour casts, low illumination, and uneven illumination. The outcomes show that our approach outperforms state-of-the-art approaches. © 2024, The Authors. All rights reserved.

关键词： Image enhancement

Multi-Level Interactive Fusion Network Based on Adversarial Learning for Fusion Classification of Hyperspectral and Lidar Data

学校读者我要写书评

暂无评论

SSRN 2024年

作者： Fan, Yingying Qian, Yurong Gong, Weijun Chu, Zhuang Qin, Yugang Muhetaer, Palidan College of Information Science and Engineering Xinjiang University Xinjiang Urumqi China College of Software Xinjiang University Xinjiang Urumqi China Key Laboratory of Signal Detection and Processing in Xinjiang Uygur Autonomous Region Xinjiang Urumqi China Key Laboratory of Software Engineering Xinjiang Urumqi China College of Information Management Xinjiang University of Finance and Economics Xinjiang Urumqi China

With the growing awareness of the limitations of unimodal data, research on joint classification usingmultimodal remote sensing data has garnered increasing attention. However, existing methods stillhave certain deficiencies in dealing with feature extraction and feature fusion of multimodal remotesensing data. In order to obtain more accurate classification results, it is crucial to fully exploit andutilize the complementary information in multimodal remote sensing data. In this paper, a multi-levelinteractive fusion network based on adversarial learning (MLIF-AL) is proposed for Hyperspectralimage (HSI) and light detection and ranging (LiDAR) data fusion classification. First, the featureextraction part based on generative adversarial network (GAN) utilizes adversarial training betweenthe generator and the discriminator to guide the network to learn more discriminative and distinguishedfeature representations. meanwhile, a cross-modal interactive information extraction (CMIIE) moduleis also designed for the generator encoding part to promote the interaction between multimodal datafrom a global perspective and realize the full utilization of multimodal complementary information. Inaddition, in the multi-level feature fusion classification (MLFFC) part, the complementarities betweenthe features learned by the neural networks at each layer are comprehensively utilized to obtain higherlevel multimodal fusion semantic information. Finally, the adversarial loss and classification loss arecombined for network training. Extensive experiments on three commonly used HSI and LiDARdatasets and comparisons with state-of-the-art methods demonstrate the effectiveness and superiorityof the model. © 2024, The Authors. All rights reserved.

关键词： Remote sensing

Superhydrophobic Pet-Cu-Ni@Mwcnt with Electropositive Layer to Enhance the Electrical Properties of Teng for Human Wearable Sensors

学校读者我要写书评

暂无评论

SSRN 2024年

作者： Yu, Shuguo Li, Peng Ding, Hanqin Shi, Fei Li, Xingang Xinjiang Key Laboratory of Solid State Physics and Devices Xinjiang University Urumqi830046 China School of Physical Science and Technology Xinjiang University Urumqi830046 China School of Computer Science and Technology Xinjiang University Urumqi830046 China Key Laboratory of Signal Detection and Processing Xinjiang University Urumqi830046 China

Flexible triboelectric nanogenerator (TENG) are ideal for sustainable and wearable electronics applications due to their excellent performance and non-toxicity. In this study, multiwalled carbon nanotubes (MWCNT) loaded plain anti-radiation shielding cloth (PET-Cu-Ni) electropositive layer (PM) is synthesized, and then assembled with PDMS to make a superhydrophobicity and flexible PM-TENG. The alternating fabric-like PET in PET-Cu-Ni increases the contact area at the tribo-electrode interface, and the bimetallic atomic Cu-Ni layer increases the electrical conductivity and electron transfer capability of triboelectric material. MWCNT surface itself has a positive charge, which promotes the generation of triboelectric charges, and the deposition of the MWCNT layer also increases the contact area of interface. MWCNT with p-type conductive forms a metal-semiconductor junction with the Cu-Ni layer, and MWCNT layer acts as a hole-transporting layer that affects triboelectric charge separation to reduces triboelectric charge loss, which greatly improves the output current of TENG. The superhydrophobic PM prevents the loss of electrostatic charge on TENG surface to the extent that it promotes stable performance output. At 5 Hz operating frequency, the short-circuit current of PM-TENG is boosted to 13.7 μA, and the open-circuit voltage and power reach 231 V and 2.8 mW, respectively. Integration of PM-TENG as a power source with the SnS2 humidity sensor constitutes TENG-driven sensor, which realizes relative humidity detection with high sensitivity and fast response recovery. In addition, a remote human-machine interface (HMI) system is constructed by integrating PM-TENG, circuit module, and APP, which realizes real-time monitoring of finger curvature and promotes the development of biomedicine. © 2024, The Authors. All rights reserved.

关键词： Multiwalled carbon nanotubes (MWCN)

Exploring Turkish Speech Recognition via Hybrid CTC/Attention Architecture and Multi-feature Fusion Network

学校读者我要写书评

暂无评论

Exploring Turkish Speech Recognition via Hybrid CTC/Attentio...

Frontiers of Electronics, Information and Computation Technologies (ICFEICT), International Conference on

作者： Nurmement Yolwas Zeyu Ren Huiru Wang Wushour Slamu Xinjiang Multilingual Information Technology Laboratory College of Information Science and Engineering Xinjiang University Urumqi China Xinjiang Key Laboratory of Signal Detection College of Information Science and Engineering Xinjiang University Urumqi China

In recent years, there has been rapid development in deep learning-based End-to-End speech recognition technology. However, the performance of Turkish speech recognition systems has been hindered by the lack of Turkish speech data. To address this, this paper focuses on several speech recognition tuning techniques. Experimental results demonstrate that the best performance is achieved when utilizing a combination of speed perturbation and noise addition for data augmentation, along with the beam search width set to 16. Furthermore, to maximize the utilization of effective feature information and enhance feature extraction accuracy, a novel feature extractor called LSPC is proposed. By combining LSPC with the LiGRU network, a shared encoder structure is formed, and model compression is achieved. The results indicate that when using only Fbank features, the performance of LSPC surpasses that of MSPC and VGGnet, resulting in a respective improvement of 1.01% and 2.53% in the word error rate (WER). Finally, building upon the aforementioned advancements, a new multifeature fusion network is proposed as the primary encoder structure. The results demonstrate that the WER of the proposed feature fusion network, based on LSPC, further improves by 0.82% and 1.94% when compared to single-feature extraction using LSPC, specifically utilizing Fbank features and spectrogram features. As a result, our model achieves comparable performance to advanced End-to-End models.

关键词：

A Speaker Recognition Method Based on Stable Learning

学校读者我要写书评

暂无评论

A Speaker Recognition Method Based on Stable Learning

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

作者： Jian Zhang Jing Ma Xiaochen Guo Lin Li Liang He School of Computer Science and Technology Xinjiang University China Xinjiang Key Laboratory of Signal Detection and Processing China School of Electronic Science and Engineering Xiamen University China Department of Electronic Engineering Tsinghua University China

With the development of deep learning, speaker recognition systems have shown increasingly better performance. The generalization ability of the models is also an important aspect of performance evaluation. Typically, a baseline system is used to compare against the improved models to demonstrate performance enhancements. However, we cannot determine the differences in learned voiceprint features between the improved models and the baseline system. This paper introduces an improved speaker recognition system based on the ECAPA-TDNN model. It utilizes stable learning to eliminate sample correlation and employs attribution analysis to compare the differences in voiceprint feature learning between the improved and baseline systems. Experimental results demonstrate that stable learning improves the model’s generalization performance and helps it learn better voiceprint features. The effectiveness and generalization capability of the proposed method are verified through experiments on the VoxCeleb, CNCeleb, and LibriSpeech datasets. This work is important for enhancing speaker recognition performance, analyzing differences in voiceprint feature learning, and promoting advancements in the field.

关键词：

Exploration on the mixed practice teaching methods of team experience under the training of new engineering talents

学校读者我要写书评

暂无评论

Exploration on the mixed practice teaching methods of team e...

Information Science and Education (ICISE-IE), International Conference on

作者： Shi Fei Jia Zhen Hong Zhou Yan Yun Chen Juan School of information science and engineering Key Laboratory of Signal Detection and Processing Xinjiang University No.666 Shengli Road Tianshan District Urumqi Xinjiang School of information science and engineering Xinjiang University No.666 Shengli Road Tianshan District Urumqi Xinjiang

Facing the requirements of the first-class discipline of electronic information engineering and the construction of new engineering courses, this paper analyzes the characteristics of practice teaching under the background of new engineering, puts forward a team experience hybrid practical teaching method for new engineering talents training, expounds the implementation process of this method in detail, and applies the method to the practice project design course of electronic information engineering experimental class. The results show that this method can improve students' self-learning ability, knowledge application ability, problem-solving ability, and engineering practice ability.

关键词： Training Knowledge engineering Information science Education Problem-solving Engineering education System analysis and design

Tablet Identification Using Ensemble Convolutional Neural Networks and Vibrational Spectroscopy

学校读者我要写书评

暂无评论

SSRN 2023年

作者： Sun, Penghui Wang, Jiajia Liu, Pei Dong, Zhilin School of Computer Science and Technology Xinjiang University Urumqi China The Key Laboratory of Signal Detection and Processing Xinjiang Uygur Autonomous Region Xinjiang University China Post-doctoral workstation of Xinjiang Xinjiang Uygur Autonomous Region Product Quality Supervision and Inspection Institute Urumqi China

Vibrational spectroscopy reflects molecular vibrational data, facilitating tablet identification. This paper introduces the DCNN-EL model, which integrates vibrational spectroscopy to accurately discern tablet categories. The DCNN-EL model is built upon the Alexnet framework, starting with the proposal of the Alexnet-BN model, enhancing model training speed and robustness by iteratively applying the BN layer in lieu of the pooling layer. Subsequently, by adjusting the parameters of Alexnet-BN, five deep convolutional neural network (DCNN) models are established and combined using an integrated machine learning approach to form the DCNN-EL model. Evaluating the DCNN-EL model with two public datasets (Tablets_Nir and Tablets_Raman) demonstrates its 100% and 91.67% accuracy, affirming its efficacy and generalizability in tablet identification. Consequently, the amalgamation of convolutional neural networks and vibrational spectroscopy bears significant practical relevance in the qualitative detection of tablets. © 2023, The Authors. All rights reserved.

关键词： Convolution