Open AccessFeature Extraction Method of Radiation Source in Deep Learning Based on Square Integral Bispectrum
Author(s) -
Yao Yao,
Yu Lu,
Yiming Chen
Publication year - 2020
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1678/1/012074
Subject(s) - computer science , bispectrum , artificial intelligence , feature extraction , pattern recognition (psychology) , deep learning , convolution (computer science) , frequency domain , artificial neural network , telecommunications , spectral density , computer vision
The feature extraction method of radiation source based on deep learning is a hotspot of specific emitter identification research. In the selection of the initial radiation source data for feature extraction, there are mainly two kinds of time series IQ data and frequency domain bispectral data. Both the IQ signal and the signal bispectrum contain the information that can characterize the fingerprint of the radiation source, and the deep learning methods mostly use different deep network structures to obtain better classification performance. This paper proposes a feature extraction method of radiation source based on bispectral data, and designs a deep network structure combining convolution and long short memory network, which has a better classification and recognition rate than a single convolution network and a single LSTM network.