Yian Zhu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Yian Zhu has a total of 92 co-authors. Here are the top co-authors by number of shared publications:
Lian Li
5 shared publications- Xianchen Shi
Beijing Institute of Computer Science and Applications, Beijing 100039, China
5 shared publications Junhua Duan
Northwestern Polytechnical University,School of Computer,Xian Shaanxi,P.R.China,710072
4 shared publicationsLixiang Zhang
4 shared publications- Lian Li
School of Software, Northwestern Polytechnical University, Xi’an, China
3 shared publications - Lin Zhang
Department of Network Security and E-Commerce, School of Computer Science, Northwestern Polytechnical University, Xi’an 710072, China
3 shared publications
Latest Publications
- 23 September 2026
- Published by Elsevier in Knowledge-Based Systems
- Vol. 352
FAME-DA: Frequency-Aware Modulated Experts with Dual Affine for multimodal named entity recognition
- 23 September 2026
- Published by Elsevier in Knowledge-Based Systems
- Vol. 352
FAME-DA: Frequency-Aware Modulated Experts with Dual Affine for multimodal named entity recognition
- 19 June 2026
- Published by World Scientific Pub Co Pte Ltd in Unmanned Systems
- p. 1-17
Fusing Constrained Autoencoder and Random Forest for Unmanned Aerial Vehicle Attacks Detection
Unmanned aerial vehicle (UAV) attacks detection constitutes a pivotal task in maintaining robust cybersecurity. However, existing approaches face difficulties in handling nonlinear and high-dimensional data, which poses a challenge to accurate attacks detection. Therefore, we propose the fusing constrained autoencoder (AE) and random forest (RF) method for UAV attacks detection. Firstly, we perform the feature compression using constrained AE, and we consider the utilization of the compressed feature to extract valuable information from data. Secondly, to enhance the distinguishability of benign and malicious samples, we learn a hyperplane for each category samples using boundary constraint loss. Thirdly, we learn the clustering center of each category samples using clustering loss to achieve inter-class discriminability and intra-class compactness, thereby enhancing model robustness. During testing, we substantiate the model feasibility and universality on UAV dataset, simulation dataset, and SensorNetGuard dataset. Especially, our method achieves 98.51% accuracy for detecting UAV sensor attacks, as for SensorNetGuard dataset, our method can realize 100% accuracy for detecting attacks.
- 19 June 2026
- Published by World Scientific Pub Co Pte Ltd in Unmanned Systems
- p. 1-17
Fusing Constrained Autoencoder and Random Forest for Unmanned Aerial Vehicle Attacks Detection
Unmanned aerial vehicle (UAV) attacks detection constitutes a pivotal task in maintaining robust cybersecurity. However, existing approaches face difficulties in handling nonlinear and high-dimensional data, which poses a challenge to accurate attacks detection. Therefore, we propose the fusing constrained autoencoder (AE) and random forest (RF) method for UAV attacks detection. Firstly, we perform the feature compression using constrained AE, and we consider the utilization of the compressed feature to extract valuable information from data. Secondly, to enhance the distinguishability of benign and malicious samples, we learn a hyperplane for each category samples using boundary constraint loss. Thirdly, we learn the clustering center of each category samples using clustering loss to achieve inter-class discriminability and intra-class compactness, thereby enhancing model robustness. During testing, we substantiate the model feasibility and universality on UAV dataset, simulation dataset, and SensorNetGuard dataset. Especially, our method achieves 98.51% accuracy for detecting UAV sensor attacks, as for SensorNetGuard dataset, our method can realize 100% accuracy for detecting attacks.
- 6 April 2026
- Published by Elsevier in Expert Systems with Applications
- Vol. 322
Adaptive gated embedding and cross-modal semantic fusion for multimodal knowledge graph completion
- 6 April 2026
- Published by Elsevier in Expert Systems with Applications
- Vol. 322
Adaptive gated embedding and cross-modal semantic fusion for multimodal knowledge graph completion
- 27 February 2026
- Published by Springer Nature in The Journal of Supercomputing
- Vol. 82 (4) , 209
Priority list scheduling algorithm for response time analysis optimization using dependency-driven non-critical nodes parallelism factor
- 27 February 2026
- Published by Springer Nature in The Journal of Supercomputing
- Vol. 82 (4) , 209
Priority list scheduling algorithm for response time analysis optimization using dependency-driven non-critical nodes parallelism factor
- Open Access
- 29 January 2026
- Published by Springer Nature in Cybersecurity
- Vol. 9 (1)
Fined-grained unmanned aerial vehicle anomaly detection by fusing multi-domain features
Unmanned aerial vehicle (UAV) fault brings about anomalies of sensors data, which can have serious and catastrophic consequences. Therefore, UAV anomaly detection (UAVAD) is crucial for its safe flight. As fail to differentiate the fined-grained feature error and fuse multi-domain features, the existing reconstruction error (RE) based anomaly detection (AD) methods are limited in detection abilities. Besides, the same RE may happen for different data. Targeting at this issue, the fined-grained UAVAD method fusing multi-domain features is proposed, which can realize UAVAD with high accuracy. Firstly, UAV sensors data is preprocessed, time-domain and frequency-domain features are extracted to form multi-domain features. Further, time-domain and frequency-domain features are integrated, and feature selection is performed on the data fused with multi-domain features. Next, the reconstruction model with double decoders is utilized to make feature extraction of UAV sensors data. Then, by considering the fined-grained feature error, under one versus others and hierarchical methods, respectively, we utilize the reconstruction model with double decoders to make AD of multi-class faults of UAV. Finally, the findings from the experiments on ALFA and simulated datasets illustrate that the proposed method works well. Especially, under hierarchical method, the proposed UAVAD method can achieve 100% accuracy for each type of UAV samples.
- 29 January 2026
- Published by Springer Nature in Cybersecurity
- Vol. 9 (1)
Open AccessFined-grained unmanned aerial vehicle anomaly detection by fusing multi-domain features
Unmanned aerial vehicle (UAV) fault brings about anomalies of sensors data, which can have serious and catastrophic consequences. Therefore, UAV anomaly detection (UAVAD) is crucial for its safe flight. As fail to differentiate the fined-grained feature error and fuse multi-domain features, the existing reconstruction error (RE) based anomaly detection (AD) methods are limited in detection abilities. Besides, the same RE may happen for different data. Targeting at this issue, the fined-grained UAVAD method fusing multi-domain features is proposed, which can realize UAVAD with high accuracy. Firstly, UAV sensors data is preprocessed, time-domain and frequency-domain features are extracted to form multi-domain features. Further, time-domain and frequency-domain features are integrated, and feature selection is performed on the data fused with multi-domain features. Next, the reconstruction model with double decoders is utilized to make feature extraction of UAV sensors data. Then, by considering the fined-grained feature error, under one versus others and hierarchical methods, respectively, we utilize the reconstruction model with double decoders to make AD of multi-class faults of UAV. Finally, the findings from the experiments on ALFA and simulated datasets illustrate that the proposed method works well. Especially, under hierarchical method, the proposed UAVAD method can achieve 100% accuracy for each type of UAV samples.
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Multimedia
- Vol. 28, 14893-14907
Diversity-Guided Content Adaptive Network for Arbitrary-Scale Video Super-Resolution
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Multimedia
- Vol. 28, 14893-14907
Diversity-Guided Content Adaptive Network for Arbitrary-Scale Video Super-Resolution
- 2 November 2025
- Published by Elsevier in Expert Systems with Applications
- Vol. 299
FFCL-KGC: Feature fusion contrastive learning via dual prompting for knowledge graph completion
- 2 November 2025
- Published by Elsevier in Expert Systems with Applications
- Vol. 299
FFCL-KGC: Feature fusion contrastive learning via dual prompting for knowledge graph completion
- Open Access
Robust Audio–Visual Speaker Localization in Noisy Aircraft Cabins for Inflight Medical Assistance
Active Speaker Localization (ASL) involves identifying both who is speaking and where they are speaking from within audiovisual content. This capability is crucial in constrained and acoustically challenging environments, such as aircraft cabins during in-flight medical emergencies. In this paper, we propose a novel end-to-end Cross-Modal Audio–Visual Fusion Network (CMAVFN) designed specifically for ASL under real-world aviation conditions, which are characterized by engine noise, dynamic lighting, occlusions from seats or oxygen masks, and frequent speaker turnover. Our model directly processes raw video frames and multi-channel ambient audio, eliminating the need for intermediate face detection pipelines. It anchors spatially resolved visual features with directional audio cues using a cross-modal attention mechanism. To enhance spatiotemporal reasoning, we introduce a dual-branch localization decoder and a cross-modal auxiliary supervision loss. Extensive experiments on public datasets (AVA-ActiveSpeaker, EasyCom) and our domain-specific AirCabin-ASL benchmark demonstrate that CMAVFN achieves robust speaker localization in noisy, occluded, and multi-speaker aviation scenarios. This framework offers a practical foundation for speech-driven interaction systems in aircraft cabins, enabling applications such as real-time crew assistance, voice-based medical documentation, and intelligent in-flight health monitoring.
Open AccessRobust Audio–Visual Speaker Localization in Noisy Aircraft Cabins for Inflight Medical Assistance
Active Speaker Localization (ASL) involves identifying both who is speaking and where they are speaking from within audiovisual content. This capability is crucial in constrained and acoustically challenging environments, such as aircraft cabins during in-flight medical emergencies. In this paper, we propose a novel end-to-end Cross-Modal Audio–Visual Fusion Network (CMAVFN) designed specifically for ASL under real-world aviation conditions, which are characterized by engine noise, dynamic lighting, occlusions from seats or oxygen masks, and frequent speaker turnover. Our model directly processes raw video frames and multi-channel ambient audio, eliminating the need for intermediate face detection pipelines. It anchors spatially resolved visual features with directional audio cues using a cross-modal attention mechanism. To enhance spatiotemporal reasoning, we introduce a dual-branch localization decoder and a cross-modal auxiliary supervision loss. Extensive experiments on public datasets (AVA-ActiveSpeaker, EasyCom) and our domain-specific AirCabin-ASL benchmark demonstrate that CMAVFN achieves robust speaker localization in noisy, occluded, and multi-speaker aviation scenarios. This framework offers a practical foundation for speech-driven interaction systems in aircraft cabins, enabling applications such as real-time crew assistance, voice-based medical documentation, and intelligent in-flight health monitoring.
- 29 April 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Consumer Electronics
- Vol. 71 (2) , 5584-5594
Implicit Multi-Scale Swin Transformer Network for Image Denoising
Image denoising has been used in various edge computing scenarios such as consumer electronics to improve the image quality and user experience. Existing image denoising methods based on Convolutional Neural Networks (CNNs) and vision Transformers achieve good performance by empirically utilizing residual neural network (ResNet) as basic component. However, ResNet lacks interpretability in network design and also long-term memory of features, potentially restricting the performance of denoising networks. In this paper, we propose an Implicit Multi-scale Swin Transformer Network (IMSNet) for image denoising, which introduces the implicit Euler scheme from the feature memory perspective. Specifically, densely connected implicit feature extraction blocks (IFEBs) are designed to learn the residual mapping between noisy and clean images. The IFEB reformulates the initial skip connection of ResNet based on the implicit Euler discretization, providing both network interpretability and long-term memory. In IFEB, the multi-scale swin Transformer block (MSB) is designed as the implicit layer to capture spatial details and non-local contextual information at different scales. Additionally, a cross-layer feature fusion block (CLFF) is proposed to further improve feature reuse capabilities. Compared to existing denoising networks, the extensive experiments demonstrate the superior performance of our IMSNet in various image denoising tasks, and the flexibility in the practical applications with proposed light model on resource-restricted platforms such as consumer electronic devices.
- 29 April 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Consumer Electronics
- Vol. 71 (2) , 5584-5594
Implicit Multi-Scale Swin Transformer Network for Image Denoising
Image denoising has been used in various edge computing scenarios such as consumer electronics to improve the image quality and user experience. Existing image denoising methods based on Convolutional Neural Networks (CNNs) and vision Transformers achieve good performance by empirically utilizing residual neural network (ResNet) as basic component. However, ResNet lacks interpretability in network design and also long-term memory of features, potentially restricting the performance of denoising networks. In this paper, we propose an Implicit Multi-scale Swin Transformer Network (IMSNet) for image denoising, which introduces the implicit Euler scheme from the feature memory perspective. Specifically, densely connected implicit feature extraction blocks (IFEBs) are designed to learn the residual mapping between noisy and clean images. The IFEB reformulates the initial skip connection of ResNet based on the implicit Euler discretization, providing both network interpretability and long-term memory. In IFEB, the multi-scale swin Transformer block (MSB) is designed as the implicit layer to capture spatial details and non-local contextual information at different scales. Additionally, a cross-layer feature fusion block (CLFF) is proposed to further improve feature reuse capabilities. Compared to existing denoising networks, the extensive experiments demonstrate the superior performance of our IMSNet in various image denoising tasks, and the flexibility in the practical applications with proposed light model on resource-restricted platforms such as consumer electronic devices.
- 13 September 2024
- Published by Springer Nature in The Journal of Supercomputing
- Vol. 80 (19) , 26751-26775
End-to-end probability analysis method for multi-core distributed systems
- 13 September 2024
- Published by Springer Nature in The Journal of Supercomputing
- Vol. 80 (19) , 26751-26775
End-to-end probability analysis method for multi-core distributed systems