Research
My research focuses on brain-aware artificial intelligence: learning useful, generalizable representations from neural and multimodal physiological signals while grounding model design in neurophysiology and cognitive science.
Major research programs
Effective Emotion Recognition from Neurophysiological Signals · 2025–present
MOE Tier-2 funded project. I serve as lead researcher for a multidisciplinary effort combining EEG, EOG, GSR, facial expression, and large multimodal models to improve emotion annotation and recognition. I oversee project execution, experimental design, model development, team supervision, and resource allocation.
EEG-Based Anxiety Regulation & Technology Translation · 2019–present
I led development of an online BCI neurofeedback system for anxiety regulation in collaboration with clinical partners. The program produced VR and non-VR prototypes, clinical deployment, and subsequent technology transfer toward a commercial digital mental-health solution. The algorithmic pipeline includes our EEG-Deformer model for mental-state decoding.
NOURISH: Next-Generation Brain-Computer-Brain Platform · 2021–2024
Within the NOURISH initiative, I led work on deep-learning algorithms for EEG-based affective and cognitive decoding. Major outcomes include LGGNet (local-global graph representations), MASA-TCN (continuous and discrete emotion decoding), and EmT (cross-subject transformer-based EEG emotion recognition).
Scent Digitalization and Computation · 2024–present
This cross-disciplinary program studies olfactory perception using EEG and multimodal sensing. I work on EEG protocol design, neural decoding, acquisition software, and analysis pipelines, together with materials-science collaborators developing conformable EEG sensors.
Generalized Mental-State Decoding from EEG · 2019–present
This program develops experiments, datasets, and learning algorithms for generalized cognitive-state decoding. It includes VR-integrated BCI data acquisition, deep neural models, technical disclosures, and patent-related work.
Selected methodological contributions
- Neurophysiology-inspired graph neural networks for EEG representation learning
- Transformer and temporal-convolution models for generalized cross-subject emotion recognition
- Cross-subject and cross-task EEG representation learning
- EEG foundation models and efficient/data-distilled large-model training
- Multimodal emotion recognition and multimodal physiological learning
- Clinical and translational BCI systems for anxiety regulation and mental health
Grants & funding contributions
- Effective Emotion Recognition from Neurophysiological Signals, MOE Tier-2 Grant (2025–2028) — funded; proposal contributor and lead researcher for execution
- PACE: Effective Prognostics and Communicative System for Disorders of Consciousness, CRP Grant — interdisciplinary proposal integrating AI, conformable BCI sensing, and clinical neuroscience
- LEEGUE: Language-aligned EEG Foundation Model for Cross-task Generalization, MOE Tier-2 Grant (2026 August Call) — proposal lead
Intellectual property
Selected patent applications and technical disclosures include Mental Arousal Level Regulation System and Method, Music-based Emotion Profiling System, System and Method for Cognitive Enhancement Based on EEG Signals, and disclosures related to EEG foundation models, data-efficient training, calibration-free BCI, multimodal emotion recognition, and EEG emotion tagging.