Classification of Infrasound Events Based on Multilevel Wavelet Transforms

EasyChair Preprint no. 10907

Abstract

Classification and recognition methods for infrasound events are widely used in various fields. Although traditional classification methods have made attempts to handle infrasound signals, there are challenges in dealing with skewed training samples in the classification model and achieving accurate classification for events with limited samples due to the rarity of certain infrasound events. To address the classification problem with a small number of chemical explosion samples in the training set, this paper proposes an improved deep convolutional neural network model for the classification of infrasound signals. In the comparative analysis, we compare the improved deep convolutional neural network with the standard LeNet and ResNet models. The experimental results demonstrate that the proposed classification model achieves similar performance to the advanced ResNet model in terms of test recognition rate, while requiring fewer covariates. This provides an advantage over previous algorithms.

Keyphrases: Convolutional Neural Network, infrasound event classification, multilevel wavelet transform

@Booklet{EasyChair:10907,
  author = {Jiang Yue and Cui YiDong},
  title = {Classification of Infrasound Events Based on Multilevel Wavelet Transforms},
  howpublished = {EasyChair Preprint no. 10907},

  year = {EasyChair, 2023}}