Open AccessResearch on the Recognition of Infrasound Signal of Nuclear Explosion by SVM and CNN
Author(s) -
Yunhui Wu,
Jiemin Zhang,
Chen Xing-min,
Shiya Zou,
Ming Yang
Publication year - 2020
Publication title -
iop conference series. earth and environmental science
Language(s) - English
Resource type - Journals
eISSN - 1755-1307
pISSN - 1755-1315
DOI - 10.1088/1755-1315/610/1/012010
Subject(s) - support vector machine , pattern recognition (psychology) , convolutional neural network , artificial intelligence , infrasound , computer science , process (computing) , signal (programming language) , speech recognition , test set , data set , set (abstract data type) , acoustics , physics , programming language , operating system
Infrasound monitoring is a nuclear test monitoring technology of the CTBTO’s international monitoring system. In order to improve the accuracy rate of atmospheric infrasound signal recognition, the recognition effectiveness of convolutional neural network(CNN) and support vector machine(SVM) are studied. According to the features of nuclear explosion infrasound, the related features are extracted from other types of signals. SVM model and CNN model are used in the experiment, and a method is designed to improve the recognition efficiency. It converts infrasound signals into images, and then uses CNN to recognize them, and the learning process is improved by combining with GAN. Compared with SVM method based on artificial design features, the experimental results show that in the case of small training data set, CNN with improved learning process can dig out the potential features of signals, and has the same recognition ability as SVM. But the recognition ability of nuclear explosion and chemical explosion is slightly better than that of SVM.