Multimedia Signal Processing and Hardware Aware Intelligent Systems Lab

Custom AI Chips & Compiler Optimization for Heterogeneous Computing Circuits

This project focuses on heterogeneous computing and intelligent chip compiler optimization for applications such as intelligent control, signal processing, tensor computing, edge perception, security-oriented computing, and autonomous driving. By integrating architecture co-design, functional verification, compiler mapping, and deployment optimization, it aims to improve design efficiency, task mapping capability, and large-scale deployment support for domestic intelligent computing platforms.

The project further develops technical methods spanning task characterization, hardware abstraction, scheduling, dataflow organization, compiler mapping, and prototype implementation, supporting efficient, secure, and application-oriented deployment on heterogeneous chip platforms.

Multimodal Physiological Signal Fusion for Early Screening of Neurological Disorders

This project focuses on early screening of neurological disorders such as Alzheimer’s disease (AD) and autism spectrum disorder (ASD) through integrated research on self-developed 3D visual stimulation devices, stereoscopic eye-movement analysis, and multimodal physiological signal fusion involving EEG, eye tracking, and related biosignals. By combining dedicated equipment development, synchronized signal acquisition, cross-modal analysis, and AI-based screening models, the project aims to build a practical and clinically meaningful multimodal screening framework that supports accurate, convenient, and scalable deployment in hospitals, communities, and home-based assessment scenarios.

AI-Driven Digital Rehabilitation for Neurological Disorders

This project focuses on multimodal rehabilitation and intelligent closed-loop management for people with stroke, mild cognitive impairment, and early Alzheimer’s disease. By integrating cognitive screening, AI-generated training content, music and video intervention, and process monitoring, it aims to support personalized, sustainable, and feedback-driven cognitive rehabilitation in community and home-based scenarios.

Building on personalized story video generation, EEG-guided music intervention, and user-management systems, the project forms an initial technical chain from content generation and process monitoring to effect evaluation and long-term follow-up.

Multimodal Affective Computing and Intelligent Analysis

This project focuses on multimodal affective computing and intelligent analysis for applications such as intelligent interaction, mental health, content understanding, and service intelligence. By modeling audio, video, text, and contextual signals, it aims to improve dynamic emotion perception, deep semantic understanding, and robust multimodal fusion in real-world scenarios.

The project builds on sustained research in multimodal representation learning, feature alignment, and robust modeling, and supports practical deployment in areas such as elderly care, postoperative monitoring, smart terminals, and related intelligent service systems.

Computational Spectral Imaging, Multi-Source Fusion, and High-Dimensional Visual Sensing

This project focuses on computational spectral imaging, multi-source image fusion, and perception-oriented analysis for high-dimensional visual sensing in complex scenarios. It integrates imaging system design, reconstruction algorithms, and application validation to improve the acquisition, fusion, and intelligent use of spatial, spectral, and cross-modal information for remote sensing, industrial inspection, and related applications.

Building on research in compressed spectral reconstruction, hyperspectral super-resolution, and multi-scale image fusion, the lab has established integrated acquisition and processing platforms that support end-to-end research from front-end sensing to back-end analysis.

AI-Driven Precision Nutrition for Chronic Diseases

This project focuses on precision nutrition assessment, dietary behavior sensing, and intelligent recommendation for health management and chronic disease intervention across the life cycle. By integrating AI, mobile internet, nutritional medicine, and multimodal health data, it aims to support personalized dietary recommendation, nutrition analysis, and digital intervention for chronic disease prevention and management.

With the Kangshi app and clinician management platform, the project combines image recognition, voice input, wearable data collection, and personal health records to enable convenient nutrition tracking, report generation, online consultation, and long-term intervention support.