Project & News
Translational Projects
Custom AI Chips & Compiler Optimization for Heterogeneous Computing Circuits
Heterogeneous processes, heterogeneous arithmetic, and deployable AI computing platforms
This project develops custom digital chips, reconfigurable accelerators, and compiler optimization methods for heterogeneous computing circuits. The key challenge is how to run increasingly complex AI workloads on heterogeneous digital computing platforms with different process technologies, dataflows, precisions, and hardware constraints.
The application direction includes edge AI chips, domain-specific inference hardware, and toolchains for efficient deployment of AI models on heterogeneous computing infrastructure.
Image placeholder
3D Stereo Visual Stimulation & Eye-Tracking Acquisition for Neurological Screening
Medical-device oriented screening for Alzheimer’s disease, MCI, and autism spectrum disorder
This project develops a self-owned 3D stereoscopic visual stimulation platform and synchronized eye-movement data acquisition system for screening neurological disorders. The system is designed to support objective, non-invasive assessment of Alzheimer’s disease, mild cognitive impairment, and autism spectrum disorder through stimulus design, gaze-response analysis, and clinically meaningful behavioral markers.
The application target is a deployable medical screening device and analysis workflow that can be extended to hospitals, rehabilitation centers, and community-health scenarios.
Multimodal Physiological Signal-Based Capability Assessment
Portable, non-invasive assessment using multimodal biosignals and AI-assisted signal modeling
This project focuses on non-invasive multimodal physiological signal analysis for human capability assessment. Representative signals include EEG, eye-tracking, EMG, heart-rate related signals, and functional near-infrared spectroscopy (fNIRS). The system emphasizes portability, practical deployment, and robust multimodal fusion for real-world evaluation scenarios.
By integrating AI-based signal generation methods, multimodal feature alignment, and physiological representation learning, the project aims to support accurate and interpretable capability evaluation based on biosignals in application-oriented settings.
Image placeholder
AI-Driven Digital Rehabilitation for Neurological Disorders
Music/video generation and multimodal audiovisual stimulation for neuro-rehabilitation
This project aims to provide digital rehabilitation therapy for patients with neurological disorders by generating adaptive music, video, and multimodal sensory stimulation through AI. The core idea is to use personalized visual and auditory content to stimulate neural pathways, support training of cognitive and emotional functions, and form a scalable intervention workflow for long-term rehabilitation.
The translational target is a practical digital therapeutics system that can be integrated into home care, rehabilitation centers, and clinical follow-up services.
Image placeholder
Multimodal Signal Analysis for Affective and Social Applications
Emotion analysis, social opinion understanding, popularity prediction, and propagation analysis
This project targets broad multimodal signal analysis by modeling audio, visual, textual, and interaction signals in real-world media content. Representative applications include emotion analysis, social opinion understanding, popularity prediction, propagation analysis, user response modeling, recommendation support, and intelligent content analytics. Short-video scenarios are a representative application case, but the methodology is not limited to short-video content.
The project is suitable for deployment in media intelligence, social-content analytics, platform governance, and public communication systems.
Image placeholder
AI-Driven Precision Nutrition for Chronic Diseases
Traditional Chinese / Western medicine integrative nutritional intervention
This project focuses on precision nutritional intervention for chronic diseases by combining AI, clinical nutrition data, behavioral monitoring, and integrative knowledge from traditional Chinese and Western medicine. It aims to support personalized nutritional recommendation, risk tracking, and intervention optimization for chronic conditions such as diabetes, obesity, and cognitive decline.
The translational direction includes digital health services, hospital-supported intervention platforms, and long-term patient-management systems.
Image placeholder
Highly Synchronous, High-Precision Multi-Source / Multi-Spectral Acquisition & Reconstruction
High-synchronization sensing, reconstruction, and domain-specific application systems
This project targets highly synchronous and high-precision multi-source / multi-spectral acquisition and reconstruction for domain-specific applications. It integrates synchronized sensing hardware, multi-band data capture, reconstruction algorithms, and task-oriented analysis to support scenarios where accurate, temporally aligned, and structurally reliable data are essential.
Potential applications include industrial inspection, intelligent monitoring, high-value target sensing, medical-assistance scenarios, and specialized perception systems.
Image placeholder