Title:
Identification of Electrical Abnormal Signals Based on Deep Learning Methods

Authors:
Ziyang Wang, China

Abstract:

In the field of abnormal event detection, optical fiber vibration data is increasingly being applied. However, existing detection methods have shortcomings, necessitating new technologies to enhance performance. This study uses optical fiber vibration data for electrical signal classification. Fiber optic sensor signals are collected under conditions such as calmness, shaking, and pressing. Through time-domain and frequency-domain feature extraction, combined with recursive feature elimination to screen key features, models such as support vector machines (SVM), linear discriminant analysis (LDA), long short-term memory networks (LSTM), gated recurrent units (GRU), and convolutional neural networks (CNN) are constructed. Experiments show that the CNN model has better classification accuracy.

DOI: http://dx.doi.org/10.51505/ijaemr.2025.1306