Prof. Hangfang Zhao, Zhejiang University (Invitation talk)
Presentation title: Generic FRI-based DOA Estimation: A Model-Fitting Method
Abstract: The bandlimited signals can be acquired through sampling and perfectly recovered from the measured samples in accordance with Shannon sampling theorem. What are less obvious are sampling schemes utilizing some sort of sparsity in the nonbandlimited signal, and this is the central theme of the finite rate of innovation (FRI). It is possible to evade Nyquist sample and recover signals exploiting sparse sampling. It is a dual problem to find the DOA estimation and the innovation rate of signals with infinite bandwidth.
Personal profile: Hang-fang Zhao received the B.E. degree in electronic engineering from Xidian University, Xi’an, China, in 1991, the M.E. degree in underwater engineering from Harbin Engineering University, Harbin, China, in 1997, and the Ph.D. degree in communication and information systems from Zhejiang University, Hangzhou, China, in 2010. From 1991 to 2012, she was with the Hangzhou Applied Acoustics Research Institute, Hangzhou, China, where she conducted research in acoustic signal processing and acoustic engineering. In 2012, she joined Zhejiang University, where she currently is a Professor at the Department of Information Science and Electronic Engineering. Her research interests include array signal processing，acoustic tomography and acoustic imaging and robust signal processing in uncertain environments.
Dr. Weichang Li, Aramco Houston Research Center, USA (Invitation talk) (On-line)
Presentation title: Deep Learning for separating surface waves from seismic reflection events
Abstract: We have developed a method to combine unsupervised and supervised deep learning for seismic ground-roll attenuation. The method consists of network components that have physics meaning and motivation, including: A. a CNN network that separates a seismic record into ground-roll and reflection signal, while minimizing the residual between the input seismic record and the sum of the generated signal and ground-roll; B. a supervised classifier that creates a maximum separation of reflection and ground-roll in the frequency-wavenumber domain; and C. a CNN network modeling ground-roll from shallow reflection events. Test results on field seismic records demonstrate the effectiveness of the proposed method in preventing signal leakage and removing ground-roll from seismic data.
|
