My research interests

I am interested in developing computational models of dynamic shape and appearance of articulated human motion, and its applications for dynamic image processing, biometrics, and human-computer interaction. In particular, my research is focused on modeling dynamic shape and appearance from video sequences for different people and different views. I proposed decomposable nonlinear generative models, which factorized a dynamic motion component and static style components, where style components are fixed in a given condition (such as person style and action type). In order to model dynamic components of visual human motions in low dimensional and intrinsic space, I utilized low dimensional manifold embedding. My long-term goal is to develop robust computational models that can decompose various types of motions, and develop human-machine interaction environments with reliable recognition of human emotion and intention.

Applications of my research are tracking human motion in arbitrary view, activity recognition and retrieval, and medical image analysis. The robust understanding of human motion in facial expressions and gestures supports human-centered computing (HCC) like affective computing, and human-robot (agent) interaction with awareness of human emotions and intensions.

Current my research interests can be grouped into modeling articulated visual human motion, processing dynamic shape and appearance image, biometrics, and human-centered computing.

Modeling articulated visual human motion(details)

• Style and content decomposition in dynamic human motion
  Representation of intrinsic body configuration in low dimensional nonlinear manifolds
• Factorized nonlinear generative models for human motion
Conceptual manifold, which is hormiomorphic to original manifold, is used for unified representation of dynamics of body configuration
• Motion primitive analysis
Motion primitive segmentation framework for human motion synthesis from motion-captured data using motion manifold learning
• Human visual preception

Processing dynamic shape and appearance image (details)

• Nonlinear active appearance models (NAAM) and facial expression analysis
  Nonlinear shape and appearance deformations with variations in different people and expressions.
• Facial expression synthesis
  We demonstrated that we can transfer dynamic facial expresison from one person to the other

Biometrics: gait recognition (details)

• Gait recognition by style analysis
  We found that high dimensional gait image sequences can be represented faithfully using LLE and other low dimensional embedding.
• Style-adaptive gait tracking
  We demonstrated simultaneous gait tracking and recognition using person-dependent global shape deformation model.

Human-centered computing (details)

• Emotion recognition: facial expression recognition
  Recognize facial expressions from video sequences
• Rehabilitation system in desktop virtual environment
  Ankle rehabilitation system using haptic device and exercise game for rehabilitation
• Gesture interface for virutual environment interaction & Sign Language Recognition and Generation
We developed gesture interface system for avatar motion control and authoring system in immersive virutal environment. Recogized sign language for communication between the deaf and normal person