Research Article
Inception Resnet V2-ECANet Based on Gramian Angular Field Image for Specific Emitter Identification
@INPROCEEDINGS{10.1007/978-3-031-30237-4_3, author={Zibo Ma and Chengyu Wu and Chen Zhong and Ao Zhan}, title={Inception Resnet V2-ECANet Based on Gramian Angular Field Image for Specific Emitter Identification}, proceedings={Machine Learning and Intelligent Communication. 7th EAI International Conference, MLICOM 2022, Virtual Event, October 23-24, 2022, Proceedings}, proceedings_a={MLICOM}, year={2023}, month={4}, keywords={Specific Emitter Identification Gramian Angular Field Convolutional Neural Network Inception Resnet V2-ECANet}, doi={10.1007/978-3-031-30237-4_3} }- Zibo Ma
Chengyu Wu
Chen Zhong
Ao Zhan
Year: 2023
Inception Resnet V2-ECANet Based on Gramian Angular Field Image for Specific Emitter Identification
MLICOM
Springer
DOI: 10.1007/978-3-031-30237-4_3
Abstract
In this paper, we seek to efficiently and accurately identify the specific emitter with the DroneRF dataset. Firstly, we convert one-dimensional data to a Gramian Angular Field (GAF) image showing a spatial correlation, and add three kinds of noise to the original GAF image to prevent overfitting. Secondly, we propose an Inception-Resnet-V2 model based on the attention mechanism ECANet, which can improve the training effect obviously. Finally, we verify the validation and accuracy of the proposed model with GAF in the experimental results. Test accuracy of Inception Resnet V2-ECANet reaches 99.5% by using the same training set of five networks through 10-fold cross-validation.
Copyright © 2022–2025 ICST
