Junsong Zhang
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
Junsong Zhang has a total of 81 co-authors. Here are the top co-authors by number of shared publications:
Rui Li
Brain Cognition and Computing Lab, National Engineering Research Center for E-Learning, Faculty of Artificial Intelligence in Education, Central China Normal University, Wuhan, China
7 shared publicationsXiaojun Qiao
Brain Cognition and Computing Lab, National Engineering Research Center of Educational Big Data, Central China Normal University, Wuhan, Hubei 430079, China
6 shared publicationsHuimin Huang
School of Education, Jianghan University, Wuhan, Hubei 430056 China
5 shared publicationsFengpeng Wang
Epilepsy Center, Xiamen Humanity Hospital, Xiamen, Fujian 361006, China; HH-SIAT Joint Laboratory for Epilepsy Research, Xiamen, Fujian 361006, China
4 shared publicationsSiyi Chen
Brain Cognition and Intelligent Computing Lab, Department of Artificial Intelligence, School of Informatics, Xiamen University, Xiamen, Fujian 361005 China
4 shared publicationsXiaobin Zhang
Epilepsy Center, Xiamen Humanity Hospital, Xiamen, Fujian 361006, China; HH-SIAT Joint Laboratory for Epilepsy Research, Xiamen, Fujian 361006, China
4 shared publications
Latest Publications
- Open Access
- 26 March 2026
- Published by Elsevier in NeuroImage
- Vol. 331, 121891
Hippocampal-parietal directed connectivity mediates brain network reconfiguration between internal and external attention: An intracranial EEG study
- 26 March 2026
- Published by Elsevier in NeuroImage
- Vol. 331, 121891
Open AccessHippocampal-parietal directed connectivity mediates brain network reconfiguration between internal and external attention: An intracranial EEG study
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Biomedical and Health Informatics
- Vol. PP, 1-14
EEGDTF: Time-Frequency Disentangled Diffusion for High-Fidelity EEG Signal Generation
EEG signal generation is hindered by challenges such as complex time-frequency structures, the lack of explicit spectral modeling, limited data availability, and limited generalization across subjects and tasks. To address these issues, we propose EEGDTF, a diffusion-based generative framework for synthesizing high-fidelity EEG signals with improved time-frequency modeling. EEGDTF first employs a multi-scale residual encoder to enhance temporal representation learning and training stability. It further introduces a dual-branch encoder-decoder architecture for time-frequency disentanglement: the frequency branch models both periodic and aperiodic components via power spectral parameterization, while the temporal branch captures waveform continuity and long-range dependencies. A frequency-guided cross-attention mechanism integrates both branches effectively. The model is optimized through a joint waveform and spectral loss, enabling stable clean-signal estimation during reverse sampling. Experiments on four benchmark datasets demonstrate that EEGDTF achieves state-of-the-art performance in both time and frequency domains, particularly under cross-subject conditions. These results underscore the model's robustness and generalizability, positioning EEGDTF as a reliable tool for EEG data augmentation and BCI-related applications.
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Biomedical and Health Informatics
- Vol. PP, 1-14
EEGDTF: Time-Frequency Disentangled Diffusion for High-Fidelity EEG Signal Generation
EEG signal generation is hindered by challenges such as complex time-frequency structures, the lack of explicit spectral modeling, limited data availability, and limited generalization across subjects and tasks. To address these issues, we propose EEGDTF, a diffusion-based generative framework for synthesizing high-fidelity EEG signals with improved time-frequency modeling. EEGDTF first employs a multi-scale residual encoder to enhance temporal representation learning and training stability. It further introduces a dual-branch encoder-decoder architecture for time-frequency disentanglement: the frequency branch models both periodic and aperiodic components via power spectral parameterization, while the temporal branch captures waveform continuity and long-range dependencies. A frequency-guided cross-attention mechanism integrates both branches effectively. The model is optimized through a joint waveform and spectral loss, enabling stable clean-signal estimation during reverse sampling. Experiments on four benchmark datasets demonstrate that EEGDTF achieves state-of-the-art performance in both time and frequency domains, particularly under cross-subject conditions. These results underscore the model's robustness and generalizability, positioning EEGDTF as a reliable tool for EEG data augmentation and BCI-related applications.
- Open Access
NeurIPT: Foundation Model for Neural Interfaces
Electroencephalography (EEG) has wide-ranging applications, from clinical diagnosis to brain-computer interfaces (BCIs). With the increasing volume and variety of EEG data, there has been growing interest in establishing foundation models (FMs) to scale up and generalize neural decoding. Despite showing early potential, applying FMs to EEG remains challenging due to substantial inter-subject, inter-task, and inter-condition variability, as well as diverse electrode configurations across recording setups. To tackle these open challenges, we propose NeurIPT, a foundation model developed for diverse EEG-based Neural Interfaces with a Pre-trained Transformer by capturing both homogeneous and heterogeneous spatio-temporal characteristics inherent in EEG signals. Temporally, we introduce Amplitude-Aware Masked Pretraining (AAMP), masking based on signal amplitude rather than random intervals, to learn robust representations across varying signal intensities beyond local interpolation. Moreover, this temporal representation is enhanced by a Progressive Mixture-of-Experts (PMoE) architecture, where specialized expert subnetworks are progressively introduced at deeper layers, adapting effectively to the diverse temporal characteristics of EEG signals. Spatially, NeurIPT leverages the 3D physical coordinates of electrodes, enabling effective transfer of embedding across varying EEG settings, and develops Intra-Inter Lobe Pooling (IILP) during fine-tuning to efficiently exploit regional brain features. Empirical evaluations across eight downstream BCI datasets, via fine-tuning, demonstrated NeurIPT consistently achieved state-of-the-art performance, highlighting its broad applicability and robust generalization. Our work pushes forward the state of FMs in EEG and offers insights into scalable and generalizable neural information processing systems.
Open AccessNeurIPT: Foundation Model for Neural Interfaces
Electroencephalography (EEG) has wide-ranging applications, from clinical diagnosis to brain-computer interfaces (BCIs). With the increasing volume and variety of EEG data, there has been growing interest in establishing foundation models (FMs) to scale up and generalize neural decoding. Despite showing early potential, applying FMs to EEG remains challenging due to substantial inter-subject, inter-task, and inter-condition variability, as well as diverse electrode configurations across recording setups. To tackle these open challenges, we propose NeurIPT, a foundation model developed for diverse EEG-based Neural Interfaces with a Pre-trained Transformer by capturing both homogeneous and heterogeneous spatio-temporal characteristics inherent in EEG signals. Temporally, we introduce Amplitude-Aware Masked Pretraining (AAMP), masking based on signal amplitude rather than random intervals, to learn robust representations across varying signal intensities beyond local interpolation. Moreover, this temporal representation is enhanced by a Progressive Mixture-of-Experts (PMoE) architecture, where specialized expert subnetworks are progressively introduced at deeper layers, adapting effectively to the diverse temporal characteristics of EEG signals. Spatially, NeurIPT leverages the 3D physical coordinates of electrodes, enabling effective transfer of embedding across varying EEG settings, and develops Intra-Inter Lobe Pooling (IILP) during fine-tuning to efficiently exploit regional brain features. Empirical evaluations across eight downstream BCI datasets, via fine-tuning, demonstrated NeurIPT consistently achieved state-of-the-art performance, highlighting its broad applicability and robust generalization. Our work pushes forward the state of FMs in EEG and offers insights into scalable and generalizable neural information processing systems.
- Open Access
- 1 May 2025
- Published by China Science Publishing & Media Ltd. in Journal of Computer-Aided Design & Computer Graphics
- Vol. 37 (5) , 780-789
Rule and Data-Driven Automatic Exhibition Hall Space Planning and Layout Generation
To address the issues of long cycles and high costs in the traditional exhibition hall layout design, we propose a method of integrating rule-based and data-driven automatic exhibition space planning and layout generation. Firstly, the specified space is divided according to the exhibition hall design rules, and the space is optimized considering beautification and connection. Then, the visual saliency of the exhibition hall space is calculated, and the position and arrangement of exhibits are selected using a priority queue. Next, using the data generated by rule-based methods as a dataset, and the layout search results from the regional center as input, we achieve the diverse generation by combining semantic prediction network and topological prediction network. Finally, through the prediction model, the optimal placement positions of exhibits are determined in decreasing order of importance. On the exhibition hall dataset, we perform comparative experiments with other methods, including HouseGAN++, Deeplayout, and DCGAN. The results indicate that the proposed method has higher visual similarity, with an average reduction of 22.64% in graph edit distance, and an average increase of 25.33% in connectivity, validating the effectiveness of this method.
- 1 May 2025
- Published by China Science Publishing & Media Ltd. in Journal of Computer-Aided Design & Computer Graphics
- Vol. 37 (5) , 780-789
Open AccessRule and Data-Driven Automatic Exhibition Hall Space Planning and Layout Generation
To address the issues of long cycles and high costs in the traditional exhibition hall layout design, we propose a method of integrating rule-based and data-driven automatic exhibition space planning and layout generation. Firstly, the specified space is divided according to the exhibition hall design rules, and the space is optimized considering beautification and connection. Then, the visual saliency of the exhibition hall space is calculated, and the position and arrangement of exhibits are selected using a priority queue. Next, using the data generated by rule-based methods as a dataset, and the layout search results from the regional center as input, we achieve the diverse generation by combining semantic prediction network and topological prediction network. Finally, through the prediction model, the optimal placement positions of exhibits are determined in decreasing order of importance. On the exhibition hall dataset, we perform comparative experiments with other methods, including HouseGAN++, Deeplayout, and DCGAN. The results indicate that the proposed method has higher visual similarity, with an average reduction of 22.64% in graph edit distance, and an average increase of 25.33% in connectivity, validating the effectiveness of this method.
- 31 March 2025
- Published by Elsevier in Neurocomputing
- Vol. 638
Improving Chinese character representation with formation tree
- 31 March 2025
- Published by Elsevier in Neurocomputing
- Vol. 638
Improving Chinese character representation with formation tree
- Open Access
- 18 December 2024
- Published by Springer Nature in Brain Informatics
- Vol. 11 (1) , 1-18
A temporal-spectral graph convolutional neural network model for EEG emotion recognition within and across subjects
EEG-based emotion recognition uses high-level information from neural activities to predict emotional responses in subjects. However, this information is sparsely distributed in frequency, time, and spatial domains and varied across subjects. To address these challenges in emotion recognition, we propose a novel neural network model named Temporal-Spectral Graph Convolutional Network (TSGCN). To capture high-level information distributed in time, spatial, and frequency domains, TSGCN considers both neural oscillation changes in different time windows and topological structures between different brain regions. Specifically, a Minimum Category Confusion (MCC) loss is used in TSGCN to reduce the inconsistencies between subjective ratings and predefined labels. In addition, to improve the generalization of TSGCN on cross-subject variation, we propose Deep and Shallow feature Dynamic Adversarial Learning (DSDAL) to calculate the distance between the source domain and the target domain. Extensive experiments were conducted on public datasets to demonstrate that TSGCN outperforms state-of-the-art methods in EEG-based emotion recognition. Ablation studies show that the mixed neural networks and our proposed methods in TSGCN significantly contribute to its high performance and robustness. Detailed investigations further provide the effectiveness of TSGCN in addressing the challenges in emotion recognition.
- 18 December 2024
- Published by Springer Nature in Brain Informatics
- Vol. 11 (1) , 1-18
Open AccessA temporal-spectral graph convolutional neural network model for EEG emotion recognition within and across subjects
EEG-based emotion recognition uses high-level information from neural activities to predict emotional responses in subjects. However, this information is sparsely distributed in frequency, time, and spatial domains and varied across subjects. To address these challenges in emotion recognition, we propose a novel neural network model named Temporal-Spectral Graph Convolutional Network (TSGCN). To capture high-level information distributed in time, spatial, and frequency domains, TSGCN considers both neural oscillation changes in different time windows and topological structures between different brain regions. Specifically, a Minimum Category Confusion (MCC) loss is used in TSGCN to reduce the inconsistencies between subjective ratings and predefined labels. In addition, to improve the generalization of TSGCN on cross-subject variation, we propose Deep and Shallow feature Dynamic Adversarial Learning (DSDAL) to calculate the distance between the source domain and the target domain. Extensive experiments were conducted on public datasets to demonstrate that TSGCN outperforms state-of-the-art methods in EEG-based emotion recognition. Ablation studies show that the mixed neural networks and our proposed methods in TSGCN significantly contribute to its high performance and robustness. Detailed investigations further provide the effectiveness of TSGCN in addressing the challenges in emotion recognition.
- 15 November 2024
- Published by Wiley in Computer Graphics Forum
- Vol. 44 (1)
Conditional Font Generation With Content Pre‐Train and Style Filter
Automatic font generation aims to streamline the design process by creating new fonts with minimal style references. This technology significantly reduces the manual labour and costs associated with traditional font design. Image‐to‐image translation has been the dominant approach, transforming font images from a source style to a target style using a few reference images. However, this framework struggles to fully decouple content from style, particularly when dealing with significant style shifts. Despite these limitations, image‐to‐image translation remains prevalent due to two main challenges faced by conditional generative models: (1) inability to handle unseen characters and (2) difficulty in providing precise content representations equivalent to the source font. Our approach tackles these issues by leveraging recent advancements in Chinese character representation research to pre‐train a robust content representation model. This model not only handles unseen characters but also generalizes to non‐existent ones, a capability absent in traditional image‐to‐image translation. We further propose a Transformer‐based Style Filter that not only accurately captures stylistic features from reference images but also handles any combination of them, fostering greater convenience for practical automated font generation applications. Additionally, we incorporate content loss with commonly used pixel‐ and perceptual‐level losses to refine the generated results from a comprehensive perspective. Extensive experiments validate the effectiveness of our method, particularly its ability to handle unseen characters, demonstrating significant performance gains over existing state‐of‐the‐art methods.
- 15 November 2024
- Published by Wiley in Computer Graphics Forum
- Vol. 44 (1)
Conditional Font Generation With Content Pre‐Train and Style Filter
Automatic font generation aims to streamline the design process by creating new fonts with minimal style references. This technology significantly reduces the manual labour and costs associated with traditional font design. Image‐to‐image translation has been the dominant approach, transforming font images from a source style to a target style using a few reference images. However, this framework struggles to fully decouple content from style, particularly when dealing with significant style shifts. Despite these limitations, image‐to‐image translation remains prevalent due to two main challenges faced by conditional generative models: (1) inability to handle unseen characters and (2) difficulty in providing precise content representations equivalent to the source font. Our approach tackles these issues by leveraging recent advancements in Chinese character representation research to pre‐train a robust content representation model. This model not only handles unseen characters but also generalizes to non‐existent ones, a capability absent in traditional image‐to‐image translation. We further propose a Transformer‐based Style Filter that not only accurately captures stylistic features from reference images but also handles any combination of them, fostering greater convenience for practical automated font generation applications. Additionally, we incorporate content loss with commonly used pixel‐ and perceptual‐level losses to refine the generated results from a comprehensive perspective. Extensive experiments validate the effectiveness of our method, particularly its ability to handle unseen characters, demonstrating significant performance gains over existing state‐of‐the‐art methods.
- Open Access
- 14 June 2024
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Applied Perception
- Vol. 21 (3) , 1-21
Color Hint-guided Ink Wash Painting Colorization with Ink Style Prediction Mechanism
We propose an end-to-end generative adversarial network that allows for controllable ink wash painting generation from sketches by specifying the colors via color hints. To the best of our knowledge, this is the first study for interactive Chinese ink wash painting colorization from sketches. To help our network understand the ink style and artistic conception, we introduced an ink style prediction mechanism for our discriminator, which enables the discriminator to accurately predict the style with the help of a pre-trained style encoder. We also designed our generator to receive multi-scale feature information from the feature pyramid network for detail reconstruction of ink wash painting. Experimental results and user study show that ink wash paintings generated by our network have higher realism and richer artistic conception than existing image generation methods.
- 14 June 2024
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Applied Perception
- Vol. 21 (3) , 1-21
Open AccessColor Hint-guided Ink Wash Painting Colorization with Ink Style Prediction Mechanism
We propose an end-to-end generative adversarial network that allows for controllable ink wash painting generation from sketches by specifying the colors via color hints. To the best of our knowledge, this is the first study for interactive Chinese ink wash painting colorization from sketches. To help our network understand the ink style and artistic conception, we introduced an ink style prediction mechanism for our discriminator, which enables the discriminator to accurately predict the style with the help of a pre-trained style encoder. We also designed our generator to receive multi-scale feature information from the feature pyramid network for detail reconstruction of ink wash painting. Experimental results and user study show that ink wash paintings generated by our network have higher realism and richer artistic conception than existing image generation methods.
- Open Access
- 7 May 2024
- Published by Elsevier in NeuroImage
- Vol. 294, 120640
Attentional control influence habituation through modulation of connectivity patterns within the prefrontal cortex: Insights from stereo-EEG
- 7 May 2024
- Published by Elsevier in NeuroImage
- Vol. 294, 120640
Open AccessAttentional control influence habituation through modulation of connectivity patterns within the prefrontal cortex: Insights from stereo-EEG