Deception Detection Using a Convolutional Neural Network
Keywords
No Thumbnail Available
Authors
Issue Date
2019-02-12
Language
en
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
During this research, I trained the convolutional neural
network called SqeeuezNet on a data set consisting of
343:503 frames. The goal was to identify what facial
regions are most expressive for lie detection. However,
due to a not well- t data set, none of the facial regions
turn out to be signi cantly important for deception de-
tection. This may also be caused by an over tted model.
Apart from that, it does seem that the area surrounding
the eyebrows and the eyes are most valuable for detect-
ing deception with each having a p value of 0:184 and
0:194 respectively. Nonetheless, di erent literature sug-
gests that there is quite high potential for arti cial neu-
ral networks, especially convolutional neural networks,
to be successful in detecting deception especially when
using multiple modalities like auditory features.
Description
Citation
Supervisor
Faculty
Faculteit der Sociale Wetenschappen