Nenglun Chen
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Nenglun Chen has a total of 20 co-authors. Here are the top co-authors by number of shared publications:
- Bangquan Liu
College of Digital Technology and Engineering, Ningbo University of Finance & Economics, Ningbo,China
1 shared publication - Dechao Sun
College of Digital Technology and Engineering, Ningbo University of Finance and Economics, Ningbo, China
1 shared publication - Feng Liang
College of Big Data and Software Engineering, Zhejiang Wanli University, Ningbo,China
1 shared publication - Kao Zhang
The School of Artificial Intelligence/School of Future Technology, Nanjing University of Information Science and Technology, Nanjing 210044, China
1 shared publication - Keke Tang
Cyberspace Institute of Advanced Technology, Guangzhou University, Guangzhou, Guangdong, China
1 shared publication - Pengyuan Quan
Nanjing University of Information Science and Technology, Nanjing, China
1 shared publication
Latest Publications
- 27 November 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 25 (1) , 2008-2018
Attentive Fusion for Efficient Wrist-Worn Gesture Recognition Based on Dual-View Cameras
Wearable hand gesture recognition has attracted considerable attention in the field of smart sensors, human-computer interaction etc. In this paper, we propose a portable device for gesture recognition based on dual-view wrist-worn cameras. Our device is easy to wear and will not impede hand movement. The main challenges we face are the sever occlusion as well as diverse lighting conditions, which significantly hinder the accuracy of gesture recognition. To mitigate these issues, we propose a novel framework based on attentive feature fusion. The overall framework mainly consists of three components, including an adaptive feature amplification module, an attentive feature enhancing as well as a cross-view attention module. Our intuition is that the detailed features from palm and dorsal hand regions can complement each other and can be regarded as strong cues for inferring hand gestures. Moreover, by adaptively adjusting the brightness of the images, the proposed method can be adaptable to diverse lighting conditions. Our method is efficient and quite effective. By comparing our work with other state-of-the-art methods, we achieve superior performance under various experiment configurations. Extensive experimental results validate that our proposed framework surpasses existing state-of-the-art works by a significant margin.
- 27 November 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 25 (1) , 2008-2018
Attentive Fusion for Efficient Wrist-Worn Gesture Recognition Based on Dual-View Cameras
Wearable hand gesture recognition has attracted considerable attention in the field of smart sensors, human-computer interaction etc. In this paper, we propose a portable device for gesture recognition based on dual-view wrist-worn cameras. Our device is easy to wear and will not impede hand movement. The main challenges we face are the sever occlusion as well as diverse lighting conditions, which significantly hinder the accuracy of gesture recognition. To mitigate these issues, we propose a novel framework based on attentive feature fusion. The overall framework mainly consists of three components, including an adaptive feature amplification module, an attentive feature enhancing as well as a cross-view attention module. Our intuition is that the detailed features from palm and dorsal hand regions can complement each other and can be regarded as strong cues for inferring hand gestures. Moreover, by adaptively adjusting the brightness of the images, the proposed method can be adaptable to diverse lighting conditions. Our method is efficient and quite effective. By comparing our work with other state-of-the-art methods, we achieve superior performance under various experiment configurations. Extensive experimental results validate that our proposed framework surpasses existing state-of-the-art works by a significant margin.
- 29 March 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Emerging Topics in Computational Intelligence
- Vol. 8 (5) , 3396-3407
Effective Single-Step Adversarial Training With Energy-Based Models
Adversarial training (AT) is one of the most effective ways against adversarial attacks. However, multi-step AT is time-consuming while single-step AT is ineffective. In this paper, we propose an Energy-AT framework to make single-step AT as effective as multi-step ones, by exploiting the two properties of energy-based models (EBM). First, we utilize the Helmholtz free energy in EBM to push generated examples to be outside of the distribution boundaries of their categories, such that they are more adversarial. Second, we apply an adaptive temperature scheme in EBM to amplify the training gradients of weak adversarial examples targetedly, such that those originally hard-to-learn examples contribute to the robustification of models also. Extensive experiments validate that Energy-AT improves the robustness of models significantly to adversarial attacks in both white-box and black-box settings, and outperforms the state-of-the-art methods.
- 29 March 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Emerging Topics in Computational Intelligence
- Vol. 8 (5) , 3396-3407
Effective Single-Step Adversarial Training With Energy-Based Models
Adversarial training (AT) is one of the most effective ways against adversarial attacks. However, multi-step AT is time-consuming while single-step AT is ineffective. In this paper, we propose an Energy-AT framework to make single-step AT as effective as multi-step ones, by exploiting the two properties of energy-based models (EBM). First, we utilize the Helmholtz free energy in EBM to push generated examples to be outside of the distribution boundaries of their categories, such that they are more adversarial. Second, we apply an adaptive temperature scheme in EBM to amplify the training gradients of weak adversarial examples targetedly, such that those originally hard-to-learn examples contribute to the robustification of models also. Extensive experiments validate that Energy-AT improves the robustness of models significantly to adversarial attacks in both white-box and black-box settings, and outperforms the state-of-the-art methods.
- Open Access
- 26 October 2023
- Published by Association for Computing Machinery (ACM)
- p. 5380-5388
Bridging Language and Geometric Primitives for Zero-shot Point Cloud Segmentation
- 26 October 2023
- Published by Association for Computing Machinery (ACM)
- p. 5380-5388
Open AccessBridging Language and Geometric Primitives for Zero-shot Point Cloud Segmentation
- 1 October 2021
- Published by Bentham Science Publishers Ltd. in Recent Advances in Computer Science and Communications
- Vol. 14 (8) , 2489-2494
Computing Salient Feature Points of 3D Model Based on Geodesic Distance and Decision Graph Clustering
Introduction:: Computing salient feature points (SFP) of 3D models has important application value in the field of computer graphics. In order to extract more effectively, a novel SFP computing algorithm based on geodesic distance and decision graph clustering is proposed. Method:: Firstly, geodesic distance of model vertices is calculated based on heat conduction equation, then average geodesic distance and importance weight of vertices are calculated. Finally, decision graph clustering method is used to calculate the decision graph of model vertices. Result and Discussion:: 3D models in SHREC 2011 dataset are selected to test the proposed algorithm. Compared with the existing algorithms, this method calculates the SFP of the 3D model from a global perspective. Results show that it is not affected by model posture and noise. Conclusion:: Our method maps the SFP of 3D model to 2D decision-making diagram, which simplifies the calculation process of SFP, improves the calculation accuracy and has strong robustness.
- 1 October 2021
- Published by Bentham Science Publishers Ltd. in Recent Advances in Computer Science and Communications
- Vol. 14 (8) , 2489-2494
Computing Salient Feature Points of 3D Model Based on Geodesic Distance and Decision Graph Clustering
Introduction:: Computing salient feature points (SFP) of 3D models has important application value in the field of computer graphics. In order to extract more effectively, a novel SFP computing algorithm based on geodesic distance and decision graph clustering is proposed. Method:: Firstly, geodesic distance of model vertices is calculated based on heat conduction equation, then average geodesic distance and importance weight of vertices are calculated. Finally, decision graph clustering method is used to calculate the decision graph of model vertices. Result and Discussion:: 3D models in SHREC 2011 dataset are selected to test the proposed algorithm. Compared with the existing algorithms, this method calculates the SFP of the 3D model from a global perspective. Results show that it is not affected by model posture and noise. Conclusion:: Our method maps the SFP of 3D model to 2D decision-making diagram, which simplifies the calculation process of SFP, improves the calculation accuracy and has strong robustness.