TinyML and edge intelligence applications in cardiovascular disease: A survey

Computers in Biology and Medicine, Journal Year: 2025, Volume and Issue: 186, P. 109653 - 109653

Published: Jan. 10, 2025

Language: Английский

---

Deep feature extraction and fault diagnosis of solenoid valves in propulsion systems based on temporal-attention GraphLSTM model

Measurement and Control, Journal Year: 2025, Volume and Issue: unknown

Published: Feb. 17, 2025

Fault diagnosis plays a crucial role in monitoring and maintaining industrial processes equipment such as discovering solenoid valve faults. Given the intricate nature of complex characterized by nonlinearity dynamics, this paper introduces novel Temporal-Attention Graph Long Short-Term Memory (TA-GraphLSTM) model for fault diagnosis, which smoothly integrates hybrid GraphLSTM network module with temporal-attention block. The architecture leverages strengths both graph LSTM neural networks to effectively handle complexities inherent data. To construct input structure data, we develop variable correlation analysis strategy based on Maximum Information Coefficient (MIC), facilitates accurate representation relationships between variables. Furthermore, incorporation allows dynamically assign weights hidden variables across time steps capture temporal dependencies effectively. proposed TA-GraphLSTM method is validated through application spacecraft propulsion system. experiment results prove effectiveness robustness model.

Language: Английский

---

Real-Time image enhancement following road scenario classification using deep learning

Measurement, Journal Year: 2025, Volume and Issue: unknown, P. 117192 - 117192

Published: March 1, 2025

Language: Английский

---

A Massive MIMO Channel Estimation Method Based on Hybrid Deep Learning Model With Regularization Techniques

International Journal of Intelligent Systems, Journal Year: 2025, Volume and Issue: 2025(1)

Published: Jan. 1, 2025

The channel estimation technique is crucial for the development of wireless communication systems. By accurately estimating state, transmission parameters such as power allocation, modulation schemes, and encoding strategies can be optimized to maximize system capacity rate. In this paper, we propose a hybrid deep learning model in multiple‐input multiple‐output (MIMO) system. combining advantages convolutions gated recurrent units (GRUs), generalization capability models across various scenarios fully utilized. Furthermore, series regularization techniques data augmentation structural complexity constraints have been introduced avoid overfitting problems. stochastic gradient descent (SGD) based on error backpropagation used iteratively train convergence. During simulation process, validated effectiveness two conditions, including quasi‐static block fading time‐varying condition. All samples are generated offline with SNRs from 10 40 dB step size 5 dB. comparison results conventional methods proven proposed method.

Language: Английский

---

Maintenance 4.0 for HVAC Systems: Addressing Implementation Challenges and Research Gaps

Smart Cities, Journal Year: 2025, Volume and Issue: 8(2), P. 66 - 66

Published: April 10, 2025

This article explores the integration of Maintenance 4.0 into HVAC (heating, ventilation, and air conditioning) systems, highlighting its essential role within framework Industry 4.0. utilizes advanced technologies such as artificial intelligence IoT sensing technologies. It also incorporates sophisticated data management techniques to transform maintenance strategies indoor ventilation systems. These innovations work together enhance energy efficiency, quality, overall system performance. The paper provides an overview various frameworks, discussing sensors in real-time monitoring environmental conditions, equipment health, consumption. highlights how AI-driven analytics, supported by data, enable predictive fault detection. Additionally, identifies key research gaps challenges that hinder widespread implementation 4.0, including issues related model interpretability, integration, scalability. proposes solutions address these challenges, techniques, explainable AI models, robust strategies, user-centered design approaches. By addressing gaps, this aims accelerate adoption contributing more sustainable, efficient, intelligent built environments.

Language: Английский

---

Chinese calligraphy character generation with component-level style learning and structure-aware guidance

Applied Soft Computing, Journal Year: 2025, Volume and Issue: unknown, P. 113159 - 113159

Published: April 1, 2025

Language: Английский

---

AI-driven innovation in emerging markets: extending the technology acceptance model–technology-organization-environment framework in small- and medium-sized enterprises

Data Science and Management, Journal Year: 2025, Volume and Issue: unknown

Published: April 1, 2025

Language: Английский

---

Prediction of PM2.5 Concentration Based on Deep Learning for High-Dimensional Time Series

Applied Sciences, Journal Year: 2024, Volume and Issue: 14(19), P. 8745 - 8745

Published: Sept. 27, 2024

PM2.5 poses a serious threat to human life and health, so the accurate prediction of concentration is essential for controlling air pollution. However, previous studies lacked generalization ability predict high-dimensional time series. Therefore, new model predicting was proposed address this in paper. Firstly, linear rectification function with leakage (LeakyRelu) used replace activation Temporal Convolutional Network (TCN) better capture dependence feature data over long distances. Next, residual structure, dilated rate, feature-matching convolution position TCN were adjusted improve performance improved (LR-TCN) reduce amount computation. Finally, (GRU-LR-TCN) established, which adaptively integrated fused Gated Recurrent Unit (GRU) LR-TCN based on inverse ratio root mean square error (RMSE) weighting. The experimental results show that, monitoring station #1001, increased RMSE, absolute (MAE), determination coefficient (R2) by 12.9%, 11.3%, 3.8%, respectively, compared baselines. Compared LR-TCN, GRU-LR-TCN index symmetric percentage (SMAPE) 7.1%. In addition, comparing estimation other models quality datasets, all indicators have advantages, it further demonstrated that exhibits superior across various proving be more efficient applicable urban concentration. This can contribute enhancing safeguarding public health.

Language: Английский

---

MA-EMD: Aligned empirical decomposition for multivariate time-series forecasting

Expert Systems with Applications, Journal Year: 2024, Volume and Issue: unknown, P. 126080 - 126080

Published: Dec. 1, 2024

Language: Английский

---

An efficient brain tumor segmentation model based on group normalization and 3D U‐Net

International Journal of Imaging Systems and Technology, Journal Year: 2024, Volume and Issue: 34(3)

Published: March 30, 2024

Abstract Accurate segmentation of brain tumors has a vital impact on clinical diagnosis and treatment, good results are helpful for the treatment this disease, which is serious threat to human health. High‐precision remains challenging task due their diverse shapes, sizes, locations, complex boundaries. Considering special structure medical tumor images, many researchers have proposed (BraTS) network based 3D U‐Net. However, there also problems such as insufficient receptive fields excessive computing costs. In paper, we propose an efficient BraTS model group normalization (GN) U‐Net (3D‐EffUNet). First, according characteristics image whole case input into model, convolution layers used extract features filter irrelevant information. Then, using main framework, convolutional module designed more precise processing features. Moreover, GN attention mechanism introduced reduce complexity without affecting performance increase awareness voxels between adjacent dimensions local space. Finally, decoder was reconstruct high‐precision The trained tested BraTS2021 dataset, experimental show that it can maintain greatly calculation cost.

Language: Английский

---