Improved semi-supervised autoencoder for deception detection

Existing algorithms of speech-based deception detection are severely restricted by the lack of sufficient number of labelled data. However, a large amount of easily available unlabelled data has not been utilized in reality. To solve this problem, this paper proposes a semi-supervised additive noise...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:PloS one Jg. 14; H. 10; S. e0223361
Hauptverfasser: Fu, Hongliang, Lei, Peizhi, Tao, Huawei, Zhao, Li, Yang, Jing
Format: Journal Article
Sprache:Englisch
Veröffentlicht: United States Public Library of Science 08.10.2019
Public Library of Science (PLoS)
Schlagworte:
Acoustics
Algorithms
Artificial Intelligence
Artificial neural networks
Biology and Life Sciences
Classification
Coders
Computer and Information Sciences
Deception
Deep learning
Engineering
Humans
Information processing
Information science
International conferences
Laboratories
Lie Detection
Neural networks
Probability distribution
Researchers
Signal processing
Social Sciences
Speech
China
ISSN:1932-6203, 1932-6203
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Existing algorithms of speech-based deception detection are severely restricted by the lack of sufficient number of labelled data. However, a large amount of easily available unlabelled data has not been utilized in reality. To solve this problem, this paper proposes a semi-supervised additive noise autoencoder model for deception detection. This model updates and optimizes the semi-supervised autoencoder and it consists of two layers of encoder and decoder, and a classifier. Firstly, it changes the activation function of the hidden layer in network according to the characteristics of the deception speech. Secondly, in order to prevent over-fitting during training, the specific ratio dropout is added at each layer cautiously. Finally, we directly connected the supervised classification task in the output of encoder to make the network more concise and efficient. Using the feature set specified by the INTERSPEECH 2009 Emotion Challenge, the experimental results on Columbia-SRI-Colorado (CSC) corpus and our own deception corpus show that the proposed model can achieve more advanced performance than other alternative methods with only a small amount of labelled data.
Bibliographie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
content type line 23
Competing Interests: The authors have declared that no competing interests exist.
ISSN:1932-6203
1932-6203
DOI:10.1371/journal.pone.0223361