Dr. Cha Zhang
Microsoft Research
Data: 17/05/2016
Horário: 14 horas
Local: Centro de Tecnologia, Bloco I - Sala 146
Bio:
Dr. Cha Zhang is a Principal Researcher in the Multimedia, Interaction and eXperience Group at Microsoft Research. He received the B.S. and M.S. degrees from Tsinghua University, Beijing, China in 1998 and 2000, respectively, both in Electronic Engineering, and the Ph.D. degree in Electrical and Computer Engineering from Carnegie Mellon University, in 2004. His current research focuses on applying various audio/image/video processing and machine learning techniques to multimedia applications, in particular, multimedia teleconferencing. Dr. Zhang has published more than 80 technical papers and holds 20+ U.S. patents. He won the best paper award at ICME 2007, the top 10% award at MMSP 2009, and the best student paper award at ICME 2010. He currently serves as an Associate Editor for IEEE Trans. on Circuits and Systems for Video Technology, and IEEE Trans. on Multimedia. He is the General Chair for ICME 2016 that will be held in Seattle, USA on July 11-15, 2016.
Sobre a Palestra:
In this presentation I will talk about graph signal processing (GSP), which has recently found many applications in social, energy, transportation, etc. In an effort to understand better its fundamental properties, we model signals on graphs as Gaussian Markov Random Fields, and present numerous important aspects of GSP, including graph construction, graph transform, graph downsampling, graph prediction, and graph-based regularization, from a probabilistic point of view. As examples, we discuss a number of methods for constructing graphs based on statistics from input data sets; we show that the graph transform is the optimal linear transform to decorrelate the signal; we describe the optimality of the Kron reduction for graph downsampling in a probabilistic sense; and we derive the optimal predictive transform coding scheme applicable to both motion prediction and intra predictive coding. Lastly, I will present one application for GSP, namely, to compress point cloud attributes where the underlining data is unstructured.
