ELLIS PhD Student · CMVS, University of Oulu & ELLIS Institute Finland
I am an ELLIS PhD student jointly affiliated with the Center for Machine Vision and Signal Analysis (CMVS), University of Oulu, and the ELLIS Institute Finland. I am supervised by Prof. Guoying Zhao, and co-supervised by Prof. Antti Honkela at the ELLIS Institute Finland and Prof. Björn Schuller within the ELLIS PhD Program. I received my M.S. in Artificial Intelligence from the Hangzhou Institute for Advanced Study, University of Chinese Academy of Sciences, in 2026, supervised by Prof. Taihao Li and Dr. Leyuan Qu.
My research mainly focuses on privacy-aware multimodal affect understanding with multimodal large language models, tracing every affective interpretation back to the audio, video and text evidence it rests on. If you are interested in collaborating, feel free to reach out via email.
Reading affect from faces, voices and language, from single frames to whole conversations.
Making perception, reasoning and adaptation more reliable when the evidence spans audio, video and text.
Protecting the people behind the data, and understanding what that protection costs.
* equal contribution
Introduces GRACE for dynamic facial emotion recognition by aligning refined linguistic cues with salient facial dynamics, achieving state of the art on DFEW, FERV39k and MAFW.
Subject-level depression detection under privacy constraints via hierarchical audio–video fusion with personalized textual cues and cross-task consistency.
Unifies global VAD intensity control with local emphasis regulation in LLM-based TTS through supervised fine-tuning and reinforcement learning, improving emotional accuracy and intensity differentiation without sacrificing synthesis quality.
A two-stage framework for disentangled, controllable talking-head synthesis; I designed the multimodal fusion module that preserves identity while sharpening affect control.
Formalizes emotion inference as a Propose–Verify–Decide protocol and uses reinforcement learning with hierarchical reward shaping to internalize abductive reasoning, producing interpretable evidence traces while outperforming strong baselines.
Extends frame-level recognition to conversation-level tracking by centering multimodal hotspots and fusing them with dialogue context, improving stability and generalization across turns.
Data Analyst, Airlines Operations Center
Lab Coordinator & Administrator
M.S. in Artificial Intelligence, Hangzhou Institute for Advanced Study
B.S. in Transportation
During my master's studies at HIAS, UCAS, I was very fortunate to be supervised by Prof. Taihao Li and Dr. Leyuan Qu, whose guidance has shaped how I think about research.
I also greatly enjoy collaborating with Haoxun Li and Hanlei Shi — I truly value our inspirational discussions and seamless teamwork.