My research direction is Analyzing human motion by sparse methods with applications to computer animation. Briefly speaking, sparse methods include those that involve vector sparse (small L0-norm) and rank sparse (low-rank).
My Research Interests range from convex optimization, computer animation, sparse coding to machine learning.
Current Research: Analyzing human motion by sparse methods with applications to computer animation
Human motion is low-rank when it is represented by a matrix. In this project, we consider the unsupervised motion completion and de-nosing problem. We show that even without using a training set as motion prior, we can still tackle this problem to some extent by exploiting the low-rank property of the human motion matrix. This is encouraging as it shows that in the process of motion capture, we don’t even need to measure the complete motion and the measurements can be noisy.
Our current work in this direction is on how to combine the rank prior with other prior such as smoothness for addressing the completion and denosing problem.
Instead of viewing the motion as whole, we break it down to poses and model the poses by sparse representation. Different from the above low-rank project, we consider the sparsity measured by $ell_0$ norm other than matrix rank.
Some preliminary outcomes our current research are:
Starting from the prevailing subspace models in modelling human motion, we propose sparse representation of poses for character posing; to our knowledge, we are the first to propose sparse coding for character posing.
Different from previous approaches, we propose to learn from the motion capture data in Euclidean space, which not only provides intuitive measurements in training error, but facilitate sparse coding and pose synthesis.
Publications / working papers/ paptent
Ranch Y. Q. Lai, Pong C. Yuen, and Kelvin K . W. Lee. Motion Capture Data Completion and Denoising by Singular Value Thresholding. Pages 45-48, Llandudno, UK , 2011. Eurographics Association (PDF,,bitex)
Ranch. Y.Q. Lai, Pong C. Yuen, K.W. Lee, J.H. Lai: Interactive character posing by sparse coding. Submitted for publication(preprint).
Ranch. Y.Q. Lai, Pong C Yuen. ProPPA: A fast algorithm for L1-minizmation and low-rank matrix completion. preprint
Ranch. Y.Q. Lai, Online Vehicle Detection for Estimating Traffic Status. preprint
J.H. Lai, F.R. Zhu, Y.Q. Lai, W. Xian, Automatic monitoring method for miner entry and exit of coal mine. Patent 200710027252.3
Ranch. Y.Q. Lai, Analyzing Human Motion By Sparse Methods, With Applications to Computer Animation. MPhil Thesis.
Last updated on May 18, 2012