Classification of Micro-Doppler Signatures Measured by Doppler Radar Through Transfer Learning

Ibrahim Alnujaim; Daegun Oh; Ikmo Park; Youngwook Kim

Classification of Micro-Doppler Signatures Measured by Doppler Radar Through Transfer Learning

Ibrahim Alnujaim, Daegun Oh, Ikmo Park, Youngwook Kim

Division of Intelligent Robot

California State University Fresno

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

5 Scopus citations

Abstract

In this paper, we investigate the feasibility of using transfer learning for the classification of micro-Doppler signatures measured by Doppler radar. A target with a non-grid body generates micro-Doppler signatures when measured by Doppler radar, which serve as an important feature for classification. However, the radar dataset is, in general, insufficient because of the high cost of its measurements. To overcome the problem of data deficiency, we propose transfer learning, which involves borrowing a classifier that has already been trained for other applications. In particular, we borrow a network trained for other micro-Doppler spectrograms rather than optical images. For the construction of the training dataset, we augment said data through generative adversarial networks. This idea is verified using human activity data measured by Doppler radar.

Original language	English
Title of host publication	13th European Conference on Antennas and Propagation, EuCAP 2019
Publisher	Institute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)	9788890701887
State	Published - Mar 2019
Event	13th European Conference on Antennas and Propagation, EuCAP 2019 - Krakow, Poland Duration: 31 Mar 2019 → 5 Apr 2019

Publication series

Name	13th European Conference on Antennas and Propagation, EuCAP 2019

Conference

Conference	13th European Conference on Antennas and Propagation, EuCAP 2019
Country/Territory	Poland
City	Krakow
Period	31/03/19 → 5/04/19

Bibliographical note

Publisher Copyright:
© 2019 European Association on Antennas and Propagation.

Keywords

Doppler processing
deep learning
generative adversarial networks
human activity classification
micro-Doppler signature
transfer learning

Cite this

@inproceedings{856180508c224a1a9a8a233b796d3314,

title = "Classification of Micro-Doppler Signatures Measured by Doppler Radar Through Transfer Learning",

abstract = "In this paper, we investigate the feasibility of using transfer learning for the classification of micro-Doppler signatures measured by Doppler radar. A target with a non-grid body generates micro-Doppler signatures when measured by Doppler radar, which serve as an important feature for classification. However, the radar dataset is, in general, insufficient because of the high cost of its measurements. To overcome the problem of data deficiency, we propose transfer learning, which involves borrowing a classifier that has already been trained for other applications. In particular, we borrow a network trained for other micro-Doppler spectrograms rather than optical images. For the construction of the training dataset, we augment said data through generative adversarial networks. This idea is verified using human activity data measured by Doppler radar.",

keywords = "Doppler processing, deep learning, generative adversarial networks, human activity classification, micro-Doppler signature, transfer learning",

author = "Ibrahim Alnujaim and Daegun Oh and Ikmo Park and Youngwook Kim",

note = "Publisher Copyright: {\textcopyright} 2019 European Association on Antennas and Propagation.; 13th European Conference on Antennas and Propagation, EuCAP 2019 ; Conference date: 31-03-2019 Through 05-04-2019",

year = "2019",

month = mar,

language = "English",

series = "13th European Conference on Antennas and Propagation, EuCAP 2019",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

booktitle = "13th European Conference on Antennas and Propagation, EuCAP 2019",

}

Alnujaim, I, Oh, D, Park, I & Kim, Y 2019, Classification of Micro-Doppler Signatures Measured by Doppler Radar Through Transfer Learning. in 13th European Conference on Antennas and Propagation, EuCAP 2019., 8740208, 13th European Conference on Antennas and Propagation, EuCAP 2019, Institute of Electrical and Electronics Engineers Inc., 13th European Conference on Antennas and Propagation, EuCAP 2019, Krakow, Poland, 31/03/19.

Classification of Micro-Doppler Signatures Measured by Doppler Radar Through Transfer Learning. / Alnujaim, Ibrahim; Oh, Daegun; Park, Ikmo et al.
13th European Conference on Antennas and Propagation, EuCAP 2019. Institute of Electrical and Electronics Engineers Inc., 2019. 8740208 (13th European Conference on Antennas and Propagation, EuCAP 2019).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Classification of Micro-Doppler Signatures Measured by Doppler Radar Through Transfer Learning

AU - Alnujaim, Ibrahim

AU - Oh, Daegun

AU - Park, Ikmo

AU - Kim, Youngwook

PY - 2019/3

Y1 - 2019/3

N2 - In this paper, we investigate the feasibility of using transfer learning for the classification of micro-Doppler signatures measured by Doppler radar. A target with a non-grid body generates micro-Doppler signatures when measured by Doppler radar, which serve as an important feature for classification. However, the radar dataset is, in general, insufficient because of the high cost of its measurements. To overcome the problem of data deficiency, we propose transfer learning, which involves borrowing a classifier that has already been trained for other applications. In particular, we borrow a network trained for other micro-Doppler spectrograms rather than optical images. For the construction of the training dataset, we augment said data through generative adversarial networks. This idea is verified using human activity data measured by Doppler radar.

AB - In this paper, we investigate the feasibility of using transfer learning for the classification of micro-Doppler signatures measured by Doppler radar. A target with a non-grid body generates micro-Doppler signatures when measured by Doppler radar, which serve as an important feature for classification. However, the radar dataset is, in general, insufficient because of the high cost of its measurements. To overcome the problem of data deficiency, we propose transfer learning, which involves borrowing a classifier that has already been trained for other applications. In particular, we borrow a network trained for other micro-Doppler spectrograms rather than optical images. For the construction of the training dataset, we augment said data through generative adversarial networks. This idea is verified using human activity data measured by Doppler radar.

KW - Doppler processing

KW - deep learning

KW - generative adversarial networks

KW - human activity classification

KW - micro-Doppler signature

KW - transfer learning

UR - http://www.scopus.com/inward/record.url?scp=85068479501&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:85068479501

T3 - 13th European Conference on Antennas and Propagation, EuCAP 2019

BT - 13th European Conference on Antennas and Propagation, EuCAP 2019

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 13th European Conference on Antennas and Propagation, EuCAP 2019

Y2 - 31 March 2019 through 5 April 2019

ER -

Classification of Micro-Doppler Signatures Measured by Doppler Radar Through Transfer Learning

Abstract

Publication series

Conference

Bibliographical note

Keywords

Other files and links

Fingerprint

Cite this