 Prof. Qunyan Ren, Institute of Acoustics, Chinese Academy of Sciences
Personal profile: Qunyan Ren, Professor in the Institute of Acoustics, Chinese Academy of Sciences. He received the B. S. degree, master degree and Doctor Degree from Harbin Engineering University in 2006, 2009, 2013, respectively. During 2010 and 2015, he has been studied and workin in U.L.B, Belgium.
He is the Chief Scientist of different national projects. His research interests include ocean acoustics, with special emphasis on geoacoustic inversion, sound propagation, and signal processing, and using AI in underwater acoustics. He has published over 100 papers, and published one book as a coauthor.
Presentation title: Two-dimensional matched filtering for echo identification with a CNN
Abstract: With strong interference of noise or reverberation, the identification of echoes from target of interest is a challenge task. An effect has been made to improve the detection rate in low single-to-noise ratio scenarios while keep false alarm rate low. In this paper, we propose a two dimensional matched filtering (2D-MF) to pre-process the recorded data from sea experiment, and then build a convolutional neural network (CNN) to identify the target echoes from background noise or reverberation. For the 2D-MF, the replica signal is divided into different sub-replicas with different frequency components, and then used to obtain 2D-MF features that can represent the amplitude frequency characteristics of the echo signals. These 2D-MF features are then exploited by the CNN for echo discrimination to determine if there are targets of interested. The method is tested on both simulated and real data, and the results from real data suggest it can improve the detection rate effectively and can keep the false alarm rate very low.
|