Journal Article

Identification of Unknown Electromagnetic Interference Sources Based on Siamese-CNN

Zhu, Feng, Xiao, Ying-Chun, Yang, Yang, Zhuang, Shengxian

2023

Overview

The prerequisite for promptly locating electromagnetic interference sources (EMIS) is the identification of EMIS. This research provides a new method for EMIS identification based on Siamese-CNN. A new convolutional neural network (CNN) structure is developed to extract the features of the EMIS. The symmetrical Siamese is adopted to enhance the number of training samples. The similarity metric of Siamese and the CNN-based subnetwork are merged in order to increase the similarity of samples from the same class and the differences between samples from different classes. A new loss function based on contrastive loss and cross-entropy loss is proposed to increase classification accuracy and discover unknown EMIS. The spectrums of EMIS are used as experimental datasets. The results show that the proposed method based on Siamese-CNN is resilient and has good generalization for training sets of various sizes. The identification accuracy for known EMIS can reach 100%, and the identification accuracy for unknown EMIS is more than 90%.

• Publisher: Springer Nature B.V
• Subject: Accuracy / Artificial neural networks / Electromagnetic interference / Similarity / Training