检索结果-内蒙古大学图书馆

Time Convolutional Network-Based Maneuvering Target Tracking with Azimuth-Doppler Measurement

SENSORS 2024年第1期24卷 263页

作者： Huang, Jianjun Hu, Haoqiang Kang, Li Shenzhen Univ Sch Intelligent Informat Proc Shenzhen 518060 Peoples R China

In the field of maneuvering target tracking, the combined observations of azimuth and Doppler may cause weak observation or non-observation in the application of traditional target-tracking algorithms. Additionally, traditional target tracking algorithms require pre-defined multiple mathematical models to accurately capture the complex motion states of targets, while model mismatch and unavoidable measurement noise lead to significant errors in target state prediction. To address those above challenges, in recent years, the target tracking algorithms based on neural networks, such as recurrent neural networks (RNNs), long short-term memory (LSTM) networks, and transformer architectures, have been widely used for their unique advantages to achieve accurate predictions. To better model the nonlinear relationship between the observation time series and the target state time series, as well as the contextual relationship among time series points, we present a deep learning algorithm called recursive downsample-convolve-interact neural network (RDCINN) based on convolutional neural network (CNN) that downsamples time series into subsequences and extracts multi-resolution features to enable the modeling of complex relationships between time series, which overcomes the shortcomings of traditional target tracking algorithms in using observation information inefficiently due to weak observation or non-observation. The experimental results show that our algorithm outperforms other existing algorithms in the scenario of strong maneuvering target tracking with the combined observations of azimuth and Doppler.

关键词： azimuth and Doppler convolutional neural network deep learning algorithm maneuvering targets tracking

来源：评论

学校读者我要写书评

暂无评论

Identify IoT Devices from Backbone Networks Using Lightweight Neural Networks 47

Identify IoT Devices from Backbone Networks Using Lightweigh...

引用

47th IEEE Conference on Local Computer Networks (LCN)

作者： Wu, Hua Fan, Xingmeng Cheng, Guang Hu, Xiaoyan Southeast Univ Sch Cyber Sci & Engn Nanjing Peoples R China Purple Mt Labs Network & Commun Secur Nanjing Peoples R China Southeast Univ Key Lab Comp Network & Informat Integrat Nanjing Peoples R China Jiangsu Prov Engn Res Ctr Secur Ubiquitous Networ Nanjing Peoples R China

ISBN: (数字)9781665480017

ISBN: (纸本)9781665480017

Due to the heterogeneity, fragmentation, and lack of visibility, Internet of Things has become the new target for attacks. Therefore, it is necessary for Internet Service Providers to identify IoT devices to prevent attacks and protect the entire network in time. In this paper, we propose an IoT device identification approach based on lightweight deep learning models using a single feature. Specifically, we analyze the traffic pattern specific to IoT devices and use one feature to characterize this pattern, reducing the time consumption. Moreover, we select multiple time scales to extract this feature for different IoT devices, achieving an accurate characterization and improving the accuracy. Furthermore, we use unidirectional flows as analysis objects, suitable for backbone networks. The evaluation results on real-world datasets show that our approach achieves an accuracy of over 99%, with one-seventeenth of the time consumption of the state-of-the-art approach, realizing the lightweight and real-time requirements.

关键词： Internet of Things devices identification specific traffic pattern deep learning algorithm

来源：评论

学校读者我要写书评

暂无评论

An active model for ranging by deep convolutional neural network and elephant herding optimization algorithm (DCNN-EHOA) in WSNs

引用

INTERNATIONAL JOURNAL OF PERVASIVE COMPUTING AND COMMUNICATIONS 2022年第2期18卷 236-249页

作者： Reddy, Adireddy Rajasekhar Rao, Appini Narayana NBKR Inst Sci & Technol Nellore India

Purpose In modern technology, the wireless sensor networks (WSNs) are generally most promising solutions for better reliability, object tracking, remote monitoring and more, which is directly related to the sensor nodes. Received signal strength indication (RSSI) is main challenges in sensor networks, which is fully depends on distance measurement. The learning algorithm based traditional models are involved in error correction, distance measurement and improve the accuracy of effectiveness. But, most of the existing models are not able to protect the user's data from the unknown or malicious data during the signal transmission. The simulation outcomes indicate that proposed methodology may reach more constant and accurate position states of the unknown nodes and the target node in WSNs domain than the existing methods. Design/methodology/approach This paper present a deep convolutional neural network (DCNN) from the adaptation of machine learning to identify the problems on deep ranging sensor networks and overthrow the problems of unknown sensor nodes localization in WSN networks by using instance parameters of elephant herding optimization (EHO) technique and which is used to optimize the localization problem. Findings In this proposed method, the signal propagation properties can be extracted automatically because of this image data and RSSI data values. Rest of this manuscript shows that the ECO can find the better performance analysis of distance estimation accuracy, localized nodes and its transmission range than those traditional algorithms. ECO has been proposed as one of the main tools to promote a transformation from unsustainable development to one of sustainable development. It will reduce the material intensity of goods and services. Originality/value The proposed technique is compared to existing systems to show the proposed method efficiency. The simulation results indicate that this proposed methodology can achieve more constant and accurate positio

关键词： Wireless sensor networks Convolutional neural network EHO algorithm deep learning algorithm Signal propagation

来源：评论

学校读者我要写书评

暂无评论

Metadata information and fundus image fusion neural network for hyperuricemia classification in diabetes

引用

COMPUTER METHODS AND PROGRAMS IN BIOMEDICINE 2024年 256卷 108382-108382页

作者： Wei, Jin Xu, Yupeng Wang, Hanying Niu, Tian Jiang, Yan Shen, Yinchen Su, Li Dou, Tianyu Peng, Yige Bi, Lei Xu, Xun Wang, Yufan Liu, Kun Shanghai Jiao Tong Univ Shanghai Engn Ctr Precise Diag & Treatment Eye Dis Shanghai Engn Ctr Visual Sci & Photomed Dept OphthalmolShanghai Gen HospSch MedNatl Cli 100 Haining Rd Shanghai 20080 Peoples R China Shanghai Jiao Tong Univ Inst Translat Med Natl Ctr Translat Med Shanghai 20080 Peoples R China Shanghai Jiao Tong Univ Shanghai Gen Hosp Dept Endocrinol & Metab Sch Med Shanghai 200080 Peoples R China

Objective: In diabetes mellitus patients, hyperuricemia may lead to the development of diabetic complications, including macrovascular and microvascular dysfunction. However, the level of blood uric acid in diabetic patients is obtained by sampling peripheral blood from the patient, which is an invasive procedure and not conducive to routine monitoring. Therefore, we developed deep learning algorithm to detect noninvasively hyperuricemia from retina photographs and metadata of patients with diabetes and evaluated performance in multiethnic populations and different subgroups. Materials and methods: To achieve the task of non-invasive detection of hyperuricemia in diabetic patients, given that blood uric acid metabolism is directly related to estimated glomerular filtration rate(eGFR), we first performed a regression task for eGFR value before the classification task for hyperuricemia and reintroduced the eGFR regression values into the baseline information. We trained 3 deep learning models: (1) metadata model adjusted for sex, age, body mass index, duration of diabetes, HbA1c, systolic blood pressure, diastolic blood pressure;(2) image model based on fundus photographs;(3)hybrid model combining image and metadata model. Data from the Shanghai General Hospital Diabetes Management Center (ShDMC) were used to develop (6091 participants with diabetes) and internally validated (using 5-fold cross-validation) the models. External testing was performed on an independent dataset (UK Biobank dataset) consisting of 9327 participants with diabetes. Results: For the regression task of eGFR, in ShDMC dataset, the coefficient of determination (R2) was 0.684+0.07 (95 % CI) for image model, 0.501+0.04 for metadata model, and 0.727+0.002 for hybrid model. In external UK Biobank dataset, a coefficient of determination (R2) was 0.647+0.06 for image model, 0.627+0.03 for metadata model, and 0.697+0.07 for hybrid model. Our method was demonstrably superior to previous methods. For the

关键词： deep learning algorithm Hyperuricemia Fundus photographs Noninvasively screening

来源：评论

学校读者我要写书评

暂无评论

STNILM: Switch Transformer based Non-Intrusive Load Monitoring for short and long duration appliances

引用

SUSTAINABLE ENERGY GRIDS & NETWORKS 2024年 37卷

作者： Varanasi, L. N. Sastry Karri, Sri Phani Krishna Natl Inst Technol Andhra Pradesh NIT AP Tadepalligudem India

Non-Intrusive Load Monitoring (NILM) is a technique used by contemporary energy management systems to predict and optimize appliance load distribution in real time. The real-time reduction of energy consumption and improvement of electricity efficiency are two major benefits of energy disaggregation. Transformer models have made NILM far better at forecasting device power values. Due to the absence of inductive bias in the local context, transformers may not be able to capture local signal patterns in sequence-to-point settings. In this work, we present a Switch Transformer based Non-Intrusive Load Monitoring (STNILM). STNILM utilizes switching and routing layers by replacing the vanilla transformer final layers to accurately estimate the power signals of short and long duration domestic appliances. It also uses self attention mechanisms to extract global dependencies between the aggregate and the domestic appliance signals. STNILM works with minimal dataset pre-processing and unbalanced. With extensive experiments and quantitative analysis, we demonstrate the efficiency and effectiveness of the proposed STNILM with considerable improvements in terms of accuracy and F1-score compared to state-of-the-art baselines.

关键词： Non-Intrusive Load Monitoring (NILM) Energy disaggregation Transformer model deep learning algorithm Sequence-to-sequence learning

来源：评论

学校读者我要写书评

暂无评论

deep learning Forecasts of Soil Moisture: Convolutional Neural Network and Gated Recurrent Unit Models Coupled with Satellite-Derived MODIS, Observations and Synoptic-Scale Climate Index Data

引用

REMOTE SENSING 2021年第4期13卷 554-554页

作者： Ahmed, A. A. Masrur Deo, Ravinesh C. Raj, Nawin Ghahramani, Afshin Feng, Qi Yin, Zhenliang Yang, Linshan Univ Southern Queensland Sch Sci Springfield Qld 4300 Australia Chinese Acad Sci Key Lab Ecohydrol Inland River Basin & Northwest Inst Ecoenvironment & Resources Lanzhou 730000 Peoples R China

Remotely sensed soil moisture forecasting through satellite-based sensors to estimate the future state of the underlying soils plays a critical role in planning and managing water resources and sustainable agricultural practices. In this paper, deep learning (DL) hybrid models (i.e., CEEMDAN-CNN-GRU) are designed for daily time-step surface soil moisture (SSM) forecasts, employing the gated recurrent unit (GRU), complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN), and convolutional neural network (CNN). To establish the objective model's viability for SSM forecasting at multi-step daily horizons, the hybrid CEEMDAN-CNN-GRU model is tested at 1st, 5th, 7th, 14th, 21st, and 30th day ahead period by assimilating a comprehensive pool of 52 predictor dataset obtained from three distinct data sources. Data comprise satellite-derived Global Land Data Assimilation System (GLDAS) repository a global, high-temporal resolution, unique terrestrial modelling system, and ground-based variables from Scientific Information Landowners (SILO) and synoptic-scale climate indices. The results demonstrate the forecasting capability of the hybrid CEEMDAN-CNN-GRU model with respect to the counterpart comparative models. This is supported by a relatively lower value of the mean absolute percentage and root mean square error. In terms of the statistical score metrics and infographics employed to test the final model's utility, the proposed CEEMDAN-CNN-GRU models are considerably superior compared to a standalone and other hybrid method tested on independent SSM data developed through feature selection approaches. Thus, the proposed approach can be successfully implemented in hydrology and agriculture management.

关键词： deep learning algorithm MODIS gated recurrent unit satellite models of soil moisture

来源：评论

学校读者我要写书评

暂无评论

MFCANN: A feature diversification framework based on local and global attention for human activity recognition

引用

ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE 2024年第PartB期133卷

作者： Yang, Zhixuan Li, Kewen Huang, Zongchao China Univ Petr East China Inst Software Coll Comp Sci & Technol 66 Changjiang West Rd Qingdao 266580 Shandong Peoples R China

Human activity recognition (HAR) is a crucial detection technique extensively employed in various contexts demanding accurate identification of human actions. Currently, mainstream HAR approaches frequently rely on sequential data generated by wearable sensors, with a primary focus on extracting feature representations that align with various human activities. Nevertheless, complex and diverse human activities encompass various kinds of features and substantial variations in the features emphasized across different activities, making it challenging for current artificial intelligence methods to achieve accurate recognition. To address this challenge, this paper introduces a network framework called the Multi-Feature Combining Attention Neural Network (MFCANN). This framework incorporates diverse feature extraction, along with both local and global feature attention mechanisms. It is composed of stacked Multi-Feature Combining Attention Blocks (MFCAB) designed by us. In contrast to traditional convolutional methods, MFCAB parallelly stacks various convolutional components, enabling the extraction of a broader range of features from human activity data. Additionally, we propose the Intra-Module Attention Block (Intra-MAB) and the Inter-Module Attention Block (Inter-MAB), which simultaneously focus on local fine-grained features within module feature maps and global distinguishing features across module feature maps. This aims to achieve more targeted feature learning, reinforcing the network's ability to distinguish between different human activities. Abundant experimental results demonstrate that the proposed MFCANN in this paper outperforms current mainstream deep learning algorithms in HAR tasks, achieving recognition accuracies of 0.9813, 0.9324, and 0.9930 on the UCI-HAR, USC-HAD, and RealWorld datasets respectively. We release our code at https://***/RangerHeart/MFCANN.

关键词： Human activity recognition Diverse feature extraction Attention mechanism deep learning algorithm Time series classification

来源：评论

学校读者我要写书评

暂无评论

Workforce forecasting in the building maintenance and repair work: Evaluating machine learning and LSTM models

引用

JOURNAL OF BUILDING ENGINEERING 2024年 95卷

作者： Cao, Nan Sing, Michael C. P. Hong Kong Polytech Univ Dept Bldg & Real Estate Kowloon Hong Kong Peoples R China

Effective workforce forecasting is critical to strategic management in construction projects, particularly ensuring staffing is optimized for efficient and timely project completion. This aspect holds particular importance in densely populated urban environments, where the complexity and scale of high-rise building construction exacerbate the need for proactive maintenance strategies. This research aims to compare the efficacy of various machine learning models, including support vector machine (SVM), extreme gradient boosting (XGBoost), gradient-boosted decision trees (GBDT), and multi-layer perceptron (MLP) models, in predicting the workforce demand for building maintenance and repair work. Accordingly, 128 sets of data composed of seven indicators collected from the Hong Kong building maintenance and repair contractors were curated to serve as input-dependent variables to forecast the demands for different labor categories in addition to the overall workforce demands. The performance metrics for the four machine learning models were evaluated using mean absolute error (MAE), mean squared error (MSE), root mean squared error (RMSE), and determination coefficient (R2), with GridSearch employed to optimize their hyperparameters for enhanced predictive accuracy. The evaluation using MAE, MSE, RMSE, and R2 revealed that the XGBoost model achieved the most favorable results for predicting workforce demand on the testing sets (0.1790, 0.0582, 0.2412, 0.9915). Compared to MLP (0.1828, 0.0775, 0.2783, 0.8821) and GBDT (0.1828, 0.0828, 0.2878, 0.8739), XGBoost demonstrated a stronger performance in this specific task. Furthermore, the study also explored the potential of a long short-term memory (LSTM) model to forecast the gross value of building maintenance and repair work undertaken by main contractors, which strongly correlates with workforce demand. This approach aimed to capture the consistent growth trend observed across various maintenance activities. Consequently,

关键词： Workforce forecasting Machine learning techniques deep learning algorithm Predictive analytics

来源：评论

学校读者我要写书评

暂无评论

Multi-hazard could exacerbate in coastal Bangladesh in the context of climate change

引用

JOURNAL OF CLEANER PRODUCTION 2024年 457卷

作者： Rahman, Mahfuzur Tian, Shufeng Tumon, Md Sakib Hasan Hossain, Md Anuwer Kim, Hyeong-Joo Islam, Md Monirul Alam, Mehtab Sadiq, Shamsher Chen, Ningsheng Ullah, Kashif Abu Zafor, Md Raju, Matiur Rahman Int Univ Business Agr & Technol IUBAT Dept Civil Engn Dhaka 1230 Bangladesh Kunsan Natl Univ Renewable Energy Res Inst 558 Daehakro Gunsan 54150 Jeollabugdo South Korea Kunsan Natl Univ Dept Civil Engn 558 Daehakro Gunsan 54150 Jeollabugdo South Korea Int Univ Business Agr & Technol IUBAT Geomat & Spatial Analyt Res Lab GSAR Dhaka 1230 Bangladesh Chinese Acad Sci Key Lab Mt Hazards & Surface Proc Inst Mt Hazards & Environm Chengdu 610041 Peoples R China Ghulam Ishaq Khan Inst Engn Sci & Technol Dept Civil Engn Swabi 23640 Pakistan Tongji Univ Coll Civil Engn Dept Geotech Engn Shanghai 200092 Peoples R China China Univ Geosci Inst Geophys & Geomat Wuhan 430074 Peoples R China Bangladesh Army Int Univ Sci & Technol BAIUST Dept Civil Engn Cumilla 3501 Bangladesh

Floods and landslides have cascading effects on coastal areas of Bangladesh. This study aims to develop multihazard maps (e.g., floods and landslides) for the coastal region, by integrating a genetic algorithm optimizer with Long Short -Term Memory deep learning algorithm (GA-LSTM-DLA), global climate model, and geospatial data. It assesses vulnerabilities to multiple hazards and projects future risks considering the Australian Community Climate and Earth System Simulator Earth System Model 1.5 (ACCESS-ESM1-5) under three climate scenarios, i.e., Shared Socioeconomic Pathways1-2.6, 2 -4.5, and 5 -8.5 (SSP1-2.6, SSP2-4.5, and SSP5-8.5). The Receiver Operating Characteristic (ROC) curve is used to assess the performance of GA-LSTM-DLA model. According to important feature obtained by random forest (RF) algorithm, most important features for flash, riverine, tidal floods, and landslides are drainage density, rainfall, and geology, respectively. The baseline susceptibility maps for flash, riverine, tidal floods, and landslides are initially recorded as 0.11%, 7.99%, 3.46%, and 0.03%. Projections for the year 2100, under different SSPs, show significant increase, i.e., SSP1-2.6 -flash 1.31%, riverine 23.74%, tidal 26.97%, landslides 2.05%;SSP2-4.5 -flash 0.12%, riverine 30.97%, tidal 10.23%, landslides 0.18%;SSP5-8.5 -flash 0.02%, riverine 17.42%, tidal 6.19%, landslides 0.14%. These projections highlight urgent need for mitigation and adaptation measures against hazards susceptibility, particularly under more extreme socioeconomic scenarios. Overall, the findings of this work are critical for policymakers to develop informed strategies for climate resilience, sustainable development, and disaster risk reduction in the coastal region of Bangladesh.

关键词： Multi-hazard deep learning algorithm Global climate model Sustainable development Coastal region

来源：评论

学校读者我要写书评

暂无评论

Discriminating BTX Molecules by the Nonselective Metal Oxide Sensor-Based Smart Sensing System

引用

ACS SENSORS 2021年第11期6卷 4167-4175页

作者： Liu, Hongyu Meng, Gang Deng, Zanhong Nagashima, Kazuki Wang, Shimao Dai, Tiantian Li, Liang Yanagida, Takeshi Fang, Xiaodong Shenzhen Technol Univ Coll New Mat & New Energies Shenzhen 518118 Peoples R China Shenzhen Univ Sch Biomed Engn Guangdong Key Lab Biomed Measurements & Ultrasoun Hlth Sci Ctr Shenzhen 518060 Peoples R China Chinese Acad Sci Anhui Inst Opt & Fine Mech Hefei 230031 Peoples R China Chinese Acad Sci Key Lab Photovolta & Energy Conservat Mat Hefei 230031 Peoples R China Univ Tokyo Grad Sch Engn Dept Appl Chem Tokyo 1138656 Japan Soochow Univ Sch Phys Sci & Technol Jiangsu Key Lab Thin Films Ctr Energy Convers Mat & Phys CECMP Suzhou 215006 Peoples R China

Discriminating structurally similar volatile organic compounds (VOCs) molecules, such as benzene, toluene, and three xylene isomers (BTX), remains a significant challenge, especially, for metal oxide semiconductor (MOS) sensors, in which selectivity is a long-standing challenge. Recent progress indicates that temperature modulation of a single MOS sensor offers a powerful route in extracting the features of adsorbed gas analytes than conventional isothermal operation. Herein, a rectangular heating waveform is applied on NiO-, WO3-, and SnO(2)(-)based sensors to gradually activate the specific gas/oxide interfacial redox reaction and generate rich (electrical) features of adsorbed BTX molecules. Upon several signal preprocessing steps, the intrinsic feature of BTX molecules can be extracted by the linear discrimination analysis (LDA) or convolutional neural network (CNN) analysis. The combination of three distinct MOS sensors noticeably benefits the recognition accuracy (with a reduced number of training iterations). Finally, a prototype of a smart BTX recognition system (including sensing electronics, sensors, Wi-Fi module, UI, PC, etc.) based on temperature modulation has been explored, which enables a prompt, accurate, and stable identification of xylene isomers in the ambient air background and raises the hope of innovating the future advanced machine olfactory system.

关键词： BTX molecules xylene isomer classification temperature modulation deep learning algorithm smart sensing system

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：