Home

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

Cui, Jinshi

Associate Professor

Research Interests:Computer Vision, developmental, psychology

Office Phone: 86-10-6275 9076

Email: cjs@cis.pku.edu.cn

Cui, Jinshi is an associate professor in the Department of Machine Intelligence, School of EECS. She obtained her B.Sc. and Ph.D. from Tsinghua University in 1999 and 2004 respectively. Her research interests include computer vision, multi-target tracking and human body posture tracking, children’s behavior analysis and its applications to and developmental psychology and pediatric. 

Dr. Cui has attended more than five research projects including NSFC, 973 programs, 863 project, etc. Her research achievements are summarized as follows:

1) Human motion tracking approaches and techniques: She addressed the fundamental problem about feature extraction from laser scan frames and videos. Based on this, she invented several novel techniques for multi-target detection and tracking using laser scan and video frames. She also invented a novel multi-target tracking framework that integrates with online learning techniques. A various attributes of targets are learned in online mode, and then used later for target identification after occlusion finishes. The work on fusion of laser and visual data is motivated from pursuing a reliable and real-time multi-target tracking system, which is difficult to achieve via only laser or only visual data.

2) Human body posture tracking approaches: Tracking generic human motion is highly challenging due to its high-dimensional state space and the various motion types involved. In order to deal with these challenges, she proposed a fusion formulation which integrates low- and high-dimensional tracking approaches into one framework. The low-dimensional approach successfully overcomes the high- dimensional problem of tracking the motions with available training data by learning motion models, but it only works with specific motion types. On the other hand, although the high-dimensional approach may recover the motions without learned models by sampling directly in the pose space, it lacks robustness and efficiency. Within the framework, the two parallel approaches, low- and high-dimensional, are fused via a probabilistic approach at each time step.

Children’s behavior analysis and its applications to developmental psychology and pediatric: According to culture-across research, higher levels of trait social anxiety have been found in East Asian countries than in Western countries. Early detection of social inhibition and supportive parenting during childhood could effectively reduce negative impact, e.g. prevention of Social Anxiety Disorder or suicidal behavior in later life. Recent progress in computer vision and eye movement tracking techniques make it possible to extract children’s behavior in real psychology research context. She has conducted extensive collaborations with psychology researchers and medical doctors. Research topics include social play behavior analysis for social withdrawal research, infants' visual attention computation for west syndrome prediction and assessment, infants visual acuity assessment, as well as autism diagnosis and assessment.