Research

UniNeuro: A Unified EEG Foundation Model for Heterogeneous EEG Task Learning

Haiyang Lu, Lianghua He, Hongzhou Chen, Xiao Chen, Wenqi Zhang

UniNeuro integrates heterogeneous-task pretraining and downstream adaptation into a single discriminative learning pipeline. A task-conditional Mixture-of-Experts with gradient reversal disentangles cross-task shared features from task-specific features. Hierarchical Prototype-Guided Contrastive Learning structures the latent space at both task and label levels. The model is pretrained on more than 10,000 hours of EEG data and evaluated across six diverse tasks.

Multimodal EEG–Text–Image Brain Decoding Model

Built approximately 200,000 paired EEG–text–image samples from 43 participants and explored cross-modal alignment and semantic decoding of brain signals.

On-device Deployment of Large EEG Models

Contributed to the Brain Science and Brain-Inspired Research initiative, studying model pruning, INT8 quantization, mixed precision, and MoE expert pruning for efficient deployment.