A New Facial Expression Recognition Scheme Based on Parallel Double Channel Convolutional Neural Network

Li, DT; Jiang, F; Qin, YB

A New Facial Expression Recognition Scheme Based on Parallel Double Channel Convolutional Neural Network

Li, DT Jiang, F

Qin, YB

Permalink

Publisher:: Springer
Publication Type:: Conference Proceeding
Citation:: Intelligent Information and Database Systems, 2020, 12034 LNAI, pp. 560-569
Issue Date:: 2020-01-01

Closed Access

	Filename	Description	Size
	Li2020_Chapter_ANewFacialExpressionRecognitio.pdf	Published version	1.37 MB		View/Open

Copyright Clearance Process

Recently Added
In Progress
Closed Access

This item is closed access and not available.

Full metadata record

Field	Value	Language
dc.contributor.author	Li, DT
dc.contributor.author	Jiang, F https://orcid.org/0000-0003-3765-561X
dc.contributor.author	Qin, YB
dc.date	2020-03-23
dc.date.accessioned	2021-05-24T07:19:22Z
dc.date.available	2021-05-24T07:19:22Z
dc.date.issued	2020-01-01
dc.identifier.citation	Intelligent Information and Database Systems, 2020, 12034 LNAI, pp. 560-569
dc.identifier.isbn	9783030420574
dc.identifier.issn	0302-9743
dc.identifier.issn	1611-3349
dc.identifier.uri	http://hdl.handle.net/10453/149159
dc.description.abstract	The conventional deep convolutional neural networks in facial expression recognition are confronted with the training inefficiency due to many layered structure with a large number of parameters, in order to cope with this challenge, in this work, an improved convolutional neural network—with parallel double channels, termed as PDC-CNN, is proposed. Within this model, we can get two different sized feature maps for the input image, and the two different sized feature maps are combined for the final recognition and judgment. In addition, in order to prevent over-fitting, we replace the traditional RuLU activation function with the Maxout model in the fully connected layer to optimize the performance of the network. We have trained and tested the new model on JAFFE dataset. Experimental results show that the proposed method can achieve 83% recognition rate, in comparison with the linear SVM, AlexNet and LeNet-5, the recognition rate of this method is improved by 14%–28%.
dc.language	en
dc.publisher	Springer
dc.relation.ispartof	Intelligent Information and Database Systems
dc.relation.ispartof	Asian Conference on Intelligent Information and Database Systems
dc.relation.ispartofseries	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
dc.relation.isbasedon	10.1007/978-3-030-42058-1_47
dc.rights	info:eu-repo/semantics/closedAccess
dc.subject.classification	Artificial Intelligence & Image Processing
dc.title	A New Facial Expression Recognition Scheme Based on Parallel Double Channel Convolutional Neural Network
dc.type	Conference Proceeding
utslib.citation.volume	12034 LNAI
utslib.location.activity	Phuket, Thailand
pubs.organisational-group	/University of Technology Sydney
pubs.organisational-group	/University of Technology Sydney/Provost
pubs.organisational-group	/University of Technology Sydney/Provost/Jumbunna
utslib.copyright.status	closed_access	*
pubs.consider-herdc	false
dc.date.updated	2021-05-24T07:19:21Z
pubs.finish-date	2020-03-26
pubs.place-of-publication	Switzerland
pubs.publication-status	Published
pubs.start-date	2020-03-23
pubs.volume	12034 LNAI
dc.location	Switzerland

Abstract:

The conventional deep convolutional neural networks in facial expression recognition are confronted with the training inefficiency due to many layered structure with a large number of parameters, in order to cope with this challenge, in this work, an improved convolutional neural network—with parallel double channels, termed as PDC-CNN, is proposed. Within this model, we can get two different sized feature maps for the input image, and the two different sized feature maps are combined for the final recognition and judgment. In addition, in order to prevent over-fitting, we replace the traditional RuLU activation function with the Maxout model in the fully connected layer to optimize the performance of the network. We have trained and tested the new model on JAFFE dataset. Experimental results show that the proposed method can achieve 83% recognition rate, in comparison with the linear SVM, AlexNet and LeNet-5, the recognition rate of this method is improved by 14%–28%.

Please use this identifier to cite or link to this item:

http://hdl.handle.net/10453/149159