Skip to Main Content
Frequently Asked Questions
Submit an ETD
Global Search Box
Need Help?
Keyword Search
Participating Institutions
Advanced Search
School Logo
Files
File List
wharton.pdf (4.55 MB)
ETD Abstract Container
Abstract Header
Deep Learning For RADAR Signal Processing
Author Info
Wharton, Michael K
ORCID® Identifier
http://orcid.org/0000-0002-6796-8732
Permalink:
http://rave.ohiolink.edu/etdc/view?acc_num=osu1619097997226646
Abstract Details
Year and Degree
2021, Master of Science, Ohio State University, Electrical and Computer Engineering.
Abstract
We address the current approaches to radar signal processing, which model radar signals with several assumptions (e.g., sparse or synchronized signals) that limit their performance and use in practical applications. We propose deep learning approaches to radar signal processing which do not make such assumptions. We present well-designed deep networks, detailed training procedures, and numerical results which show our deep networks outperform current approaches. In the first part of this thesis, we consider synthetic aperture radar (SAR) image recovery and classification from sub-Nyquist samples, i.e., compressive SAR. Our approach is to first apply back-projection and then use a deep convolutional neural network (CNN) to de-alias the result. Importantly, our CNN is trained to be agnostic to the subsampling pattern. Relative to the basis pursuit (i.e., sparsity-based) approach to compressive SAR recovery, our CNN-based approach is faster and more accurate, in terms of both image recovery MSE and downstream classification accuracy, on the MSTAR dataset. In the second part of this thesis, we consider the problem of classifying multiple overlapping phase-modulated radar waveforms given raw signal data. To do this, we design a complex-valued residual deep neural network and apply data augmentations during training to make our network robust to time synchronization, pulse width, and SNR. We demonstrate that our optimized network significantly outperforms the current state-of-the-art in terms of classification accuracy, especially in the asynchronous setting.
Committee
Philip Schniter (Advisor)
Emre Ertin (Committee Member)
Pages
46 p.
Subject Headings
Electrical Engineering
;
Engineering
Keywords
Deep learning, deep neural networks, radar, signal processing, compressive SAR, automatic target recognition, cognitive radar, classification, multi-label classification
Recommended Citations
Refworks
EndNote
RIS
Mendeley
Citations
Wharton, M. K. (2021).
Deep Learning For RADAR Signal Processing
[Master's thesis, Ohio State University]. OhioLINK Electronic Theses and Dissertations Center. http://rave.ohiolink.edu/etdc/view?acc_num=osu1619097997226646
APA Style (7th edition)
Wharton, Michael.
Deep Learning For RADAR Signal Processing.
2021. Ohio State University, Master's thesis.
OhioLINK Electronic Theses and Dissertations Center
, http://rave.ohiolink.edu/etdc/view?acc_num=osu1619097997226646.
MLA Style (8th edition)
Wharton, Michael. "Deep Learning For RADAR Signal Processing." Master's thesis, Ohio State University, 2021. http://rave.ohiolink.edu/etdc/view?acc_num=osu1619097997226646
Chicago Manual of Style (17th edition)
Abstract Footer
Document number:
osu1619097997226646
Download Count:
390
Copyright Info
© 2021, all rights reserved.
This open access ETD is published by The Ohio State University and OhioLINK.
Release 3.2.12