Please use this identifier to cite or link to this item: https://hdl.handle.net/10356/78870
Title: Audio and video-based extremists detection
Authors: Xu, Sijia
Keywords: Engineering::Electrical and electronic engineering
Issue Date: 2019
Abstract: Since the 21st century, terrorist crimes have spread around the world, which poses a severe threat to the peace and security of the international community, especially in the unlimited Internet. People wish that the extremist can be detected by the autonomous algorithm to avoid the propagation of extremist video and pictures. This dissertation proposes a method based on machine learning to identify if there are any extremists in random videos on the Internet. 1140 videos, with 660 extremist videos and 480 non-extremist videos are utilized in this project as the database for feature extraction and algorithm implementation. In the first step, the input videos are pre-processed into suitable formats before the feature extraction. In the second step, two feature extraction methods (OpenSMILE and Affectiva) are exploited to extract the emotion features hidden in the origin videos. In the third step, five machine learning algorithms (Random Forrest, SVM with RBF kernel, SVM with linear kernel, AdaBoost and Decision Tree) are used in the extremist evaluation base on the extracted features in the second step. At last, the results of extremist detection of each strategy are compared by four evaluation standards in machine learning (Precision, Recall, F-score and Accuracy), and SVM with RBF kernel which has satisfactory result in evaluation is chosen for the detection method with both OpenSMILE and Affectiva applying on the original videos at the same time.
URI: http://hdl.handle.net/10356/78870
Fulltext Permission: restricted
Fulltext Availability: With Fulltext
Appears in Collections:EEE Theses

Files in This Item:
File Description SizeFormat 
Xu_Sijia_MSc_Thesis_Soft_Copy JD.pdf
  Restricted Access
Main article2.66 MBAdobe PDFView/Open

Page view(s)

104
Updated on May 7, 2021

Download(s)

6
Updated on May 7, 2021

Google ScholarTM

Check

Items in DR-NTU are protected by copyright, with all rights reserved, unless otherwise indicated.