Radar Emitter Identification Based on Deep Convolutional Neural Network

Aiming at the identification and classification of radar radiation sources, this paper proposes a classification method based on the Convolutional Neural Network(CNN) for radar signal classification. Firstly, this paper sets the appropriate learning rate, batch size, iteration number, momentum and w...

Full description

Saved in:
Bibliographic Details
Published in2018 International Conference on Control, Automation and Information Sciences (ICCAIS) pp. 309 - 314
Main Authors Kong, Mingxin, Zhang, Jing, Liu, Weifeng, Zhang, Guilin
Format Conference Proceeding
LanguageEnglish
Published IEEE 01.10.2018
Subjects
CNN
Deep Learning
Feature extraction
Feature Representation
Neural networks
Radar
Radar Emitter Identification
Spectrogram
STFT
Support vector machines
Time-frequency analysis
Online AccessGet full text

Cover

More Information
Summary:Aiming at the identification and classification of radar radiation sources, this paper proposes a classification method based on the Convolutional Neural Network(CNN) for radar signal classification. Firstly, this paper sets the appropriate learning rate, batch size, iteration number, momentum and weight decay coefficient. Secondly, the time domain real part waveform signal is modeled and the network structure is selected for analysis. Finally, according to the spectrogram of the time domain waveform by Short Time Fourier Transform(STFT), design two different convolutional neural network models. The results show that the network learns more distinguishing feature representations and has better generalization capabilities after STFT. The Deep Learning of CNN has a greater advantage in extracting the feature representations in the spectrogram of the radar signal.
ISSN:2475-7896
DOI:10.1109/ICCAIS.2018.8570480