Research研究方向
Ubiquitous Psychological Computing and Embodied Emotional Intelligence as two core research frameworks for human-centered AI.以普适心理计算与具身情感智能两大框架为核心，构建更理解人的智能系统。
Research Mission: Affective AI in the Real World研究使命：面向真实世界的情感智能
MAC-Lab's name comes from Multimedia Affective Computing, the lab's original research foundation. From multimodal emotion perception and affective interaction, the agenda has extended toward Ubiquitous Psychological Computing and Embodied Emotional Intelligence: two connected lines that move from affective signals to mental-state modeling, and from virtual models to embodied interaction.MAC-Lab 的名字来自 Multimedia Affective Computing，即多模态情感计算，这是实验室最初的研究基础。从多模态情绪感知和情感交互出发，团队进一步延伸到普适心理计算与具身情感智能，两条主线分别连接情感信号与心理状态建模、虚拟模型与具身交互系统。
A Long-Term Route, Not a Hotspot Portfolio长期坚持方向，而不是追逐热点
The lab's route grows from Professor Sun's doctoral training at Dalian University of Technology and Tokushima University, and from years of work in natural language processing, human-computer interaction, and multimodal affective computing. Large models, digital humans, smart cockpits, and embodied agents are treated as new carriers for the same long-term question: how can AI better understand and support people?实验室的研究路线，来自孙晓教授在大连理工大学和日本国立德岛大学的中日双博士训练，也来自团队多年在自然语言处理、人机交互和多模态情感计算方向的积累。大模型、数字人、智能座舱和具身智能不是孤立热点，而是同一个长期问题的新载体：AI 如何更好地理解和支持人。
Ubiquitous Psychological Computing普适心理计算框架
This framework extends affective computing from isolated emotion recognition to continuous psychological modeling in pervasive scenarios. It treats emotion as a dynamic process shaped by intensity, time, environment, personality, physiological-behavioral signals, and group context, enabling assessment, early warning, and personalized intervention.该框架把情感计算从孤立的情绪识别推进到普适场景中的连续心理建模：从情绪反应程度、时间、环境、个性、生理行为信号和群体情境等维度理解复杂情感变化，支撑心理评估、风险预警和个性化干预。
Embodied Emotional Intelligence具身情感智能框架
This framework brings affective computing into embodied agents. It connects multimodal emotional perception, affective understanding, expressive generation, and situated interaction so that robots, digital humans, smart cockpits, eldercare companions, and active-health systems can move from task execution to trusted emotional interaction.该框架把情感计算带入具身智能体：连接多模态情感感知、情感理解、情感表达生成与情境化交互，让机器人、数字人、智能座舱、养老陪伴和主动健康系统从“能执行任务”走向“能建立可信情感互动”。
Affective NLP and Emotional Dialogue语言智能与情感对话
We develop models for Chinese language understanding, affective semantics, emotional conversation generation, empathetic response, personality-aware dialogue, and LLM-based emotional support. The goal is to help AI understand not only what is said, but also the emotional intent, social relationship, and psychological need behind the words.研究中文语言理解、情感语义、人格建模、情感支持对话和多智能体协作，让大模型不仅会回答，更能理解“为什么这样说”。
Multimodal Affective Computing多模态情绪与心理状态建模
We integrate text, speech, facial expression, micro-expression, gait, posture, eye movement, physiological signals, and environmental context. Research topics include complex emotion recognition, humor detection, stress estimation, depression and anxiety detection, cognitive load recognition, and temporal affect dynamics.融合文本、语音、面部表情、微表情、姿态、步态、眼动和生理信号，构建面向复杂情绪、心理压力、认知负荷和抑郁焦虑的识别模型。
Ubiquitous Psychological Computing普适心理计算
We propose and advance ubiquitous psychological computing: a framework for sensing, profiling, assessing, and supporting mental states in everyday scenarios. This direction emphasizes non-contact sensing, dynamic psychological modeling, long-term state tracking, early warning, and human-centered intervention.把心理状态感知从实验室带到真实空间，面向家庭、学校、医院、座舱和特种作业环境，形成非接触评测、动态画像、风险预警与干预闭环。
Embodied Emotional Intelligence具身情感智能
We study emotional interaction models and embodied agent architectures that act as an emotional brain for physical robots, digital humans, smart devices, and human-machine coexistence environments.研究具身情感交互大模型和智能体架构，为机器人、数字人、智能终端和人机共生环境提供可部署的“情感大脑”。