Unsupervised feature learning for EEG-based emotion recognition

Zirui Lan, Olga Sourina, Lipo Wang, Reinhold Scherer, Gernot Müller-Putz

Research output: Chapter in Book/Report/Conference proceedingConference contributionResearchpeer-review

Abstract

Spectral band power features are one of the most widely used features in the studies of electroencephalogram (EEG)-based emotion recognition. The power spectral density of EEG signals is partitioned into different bands such as delta, theta, alpha and beta band etc. Though based on neuroscientific findings, the partition of frequency bands is somewhat on an ad-hoc basis, and the definition of frequency ranges of the bands of interest can vary between studies. On the other hand, it is also arguable that one definition of power bands could perform equally well on all subjects. In this paper, we propose to use autoencoder to automatically learn from each subject the salient frequency components from power spectral density estimated as periodogram by Fast Fourier Transform (FFT). We propose a network architecture especially for EEG feature extraction, one that adopts hidden unit clustering with added pooling neuron per cluster. The classification accuracy with features extracted by our proposed method is benchmarked against that with standard power features. Experimental results show that our proposed feature extraction method achieves accuracy ranging from 44% to 59% for three-emotion classification. We also see a 4-20% accuracy improvement over standard band power features.

Original languageEnglish
Title of host publicationProceedings - 2017 International Conference on Cyberworlds, CW 2017 - in cooperation with
Subtitle of host publicationEurographics Association International Federation for Information Processing ACM SIGGRAPH
PublisherInstitute of Electrical and Electronics Engineers
Pages182-185
Number of pages4
Volume2017-January
ISBN (Electronic)9781538620892
DOIs
Publication statusPublished - 27 Nov 2017
Event2017 International Conference on Cyberworlds, CW 2017 - Chester, United Kingdom
Duration: 20 Sep 201722 Sep 2017

Conference

Conference2017 International Conference on Cyberworlds, CW 2017
CountryUnited Kingdom
CityChester
Period20/09/1722/09/17

Fingerprint

Emotion Recognition
Electroencephalography
Power spectral density
Feature extraction
Power Spectral Density
Network architecture
Fast Fourier transforms
Frequency bands
Neurons
Feature Extraction
Periodogram
Pooling
Fast Fourier transform
Network Architecture
Learning
Electroencephalogram
Neuron
Partition
Clustering
Vary

Keywords

  • Autoencoder
  • Brain-computer-interface (BCI)
  • Electroencephalogram (EEG)
  • Emotion classification
  • Power spectral density
  • Unsupervised feature extraction

ASJC Scopus subject areas

  • Artificial Intelligence
  • Computer Networks and Communications
  • Computer Science Applications
  • Media Technology
  • Modelling and Simulation

Cite this

Lan, Z., Sourina, O., Wang, L., Scherer, R., & Müller-Putz, G. (2017). Unsupervised feature learning for EEG-based emotion recognition. In Proceedings - 2017 International Conference on Cyberworlds, CW 2017 - in cooperation with: Eurographics Association International Federation for Information Processing ACM SIGGRAPH (Vol. 2017-January, pp. 182-185). Institute of Electrical and Electronics Engineers. https://doi.org/10.1109/CW.2017.19

Unsupervised feature learning for EEG-based emotion recognition. / Lan, Zirui; Sourina, Olga; Wang, Lipo; Scherer, Reinhold; Müller-Putz, Gernot.

Proceedings - 2017 International Conference on Cyberworlds, CW 2017 - in cooperation with: Eurographics Association International Federation for Information Processing ACM SIGGRAPH. Vol. 2017-January Institute of Electrical and Electronics Engineers, 2017. p. 182-185.

Research output: Chapter in Book/Report/Conference proceedingConference contributionResearchpeer-review

Lan, Z, Sourina, O, Wang, L, Scherer, R & Müller-Putz, G 2017, Unsupervised feature learning for EEG-based emotion recognition. in Proceedings - 2017 International Conference on Cyberworlds, CW 2017 - in cooperation with: Eurographics Association International Federation for Information Processing ACM SIGGRAPH. vol. 2017-January, Institute of Electrical and Electronics Engineers, pp. 182-185, 2017 International Conference on Cyberworlds, CW 2017, Chester, United Kingdom, 20/09/17. https://doi.org/10.1109/CW.2017.19
Lan Z, Sourina O, Wang L, Scherer R, Müller-Putz G. Unsupervised feature learning for EEG-based emotion recognition. In Proceedings - 2017 International Conference on Cyberworlds, CW 2017 - in cooperation with: Eurographics Association International Federation for Information Processing ACM SIGGRAPH. Vol. 2017-January. Institute of Electrical and Electronics Engineers. 2017. p. 182-185 https://doi.org/10.1109/CW.2017.19
Lan, Zirui ; Sourina, Olga ; Wang, Lipo ; Scherer, Reinhold ; Müller-Putz, Gernot. / Unsupervised feature learning for EEG-based emotion recognition. Proceedings - 2017 International Conference on Cyberworlds, CW 2017 - in cooperation with: Eurographics Association International Federation for Information Processing ACM SIGGRAPH. Vol. 2017-January Institute of Electrical and Electronics Engineers, 2017. pp. 182-185
@inproceedings{bb4b13c381de4dd7ad49ec6a5049018c,
title = "Unsupervised feature learning for EEG-based emotion recognition",
abstract = "Spectral band power features are one of the most widely used features in the studies of electroencephalogram (EEG)-based emotion recognition. The power spectral density of EEG signals is partitioned into different bands such as delta, theta, alpha and beta band etc. Though based on neuroscientific findings, the partition of frequency bands is somewhat on an ad-hoc basis, and the definition of frequency ranges of the bands of interest can vary between studies. On the other hand, it is also arguable that one definition of power bands could perform equally well on all subjects. In this paper, we propose to use autoencoder to automatically learn from each subject the salient frequency components from power spectral density estimated as periodogram by Fast Fourier Transform (FFT). We propose a network architecture especially for EEG feature extraction, one that adopts hidden unit clustering with added pooling neuron per cluster. The classification accuracy with features extracted by our proposed method is benchmarked against that with standard power features. Experimental results show that our proposed feature extraction method achieves accuracy ranging from 44{\%} to 59{\%} for three-emotion classification. We also see a 4-20{\%} accuracy improvement over standard band power features.",
keywords = "Autoencoder, Brain-computer-interface (BCI), Electroencephalogram (EEG), Emotion classification, Power spectral density, Unsupervised feature extraction",
author = "Zirui Lan and Olga Sourina and Lipo Wang and Reinhold Scherer and Gernot M{\"u}ller-Putz",
year = "2017",
month = "11",
day = "27",
doi = "10.1109/CW.2017.19",
language = "English",
volume = "2017-January",
pages = "182--185",
booktitle = "Proceedings - 2017 International Conference on Cyberworlds, CW 2017 - in cooperation with",
publisher = "Institute of Electrical and Electronics Engineers",
address = "United States",

}

TY - GEN

T1 - Unsupervised feature learning for EEG-based emotion recognition

AU - Lan, Zirui

AU - Sourina, Olga

AU - Wang, Lipo

AU - Scherer, Reinhold

AU - Müller-Putz, Gernot

PY - 2017/11/27

Y1 - 2017/11/27

N2 - Spectral band power features are one of the most widely used features in the studies of electroencephalogram (EEG)-based emotion recognition. The power spectral density of EEG signals is partitioned into different bands such as delta, theta, alpha and beta band etc. Though based on neuroscientific findings, the partition of frequency bands is somewhat on an ad-hoc basis, and the definition of frequency ranges of the bands of interest can vary between studies. On the other hand, it is also arguable that one definition of power bands could perform equally well on all subjects. In this paper, we propose to use autoencoder to automatically learn from each subject the salient frequency components from power spectral density estimated as periodogram by Fast Fourier Transform (FFT). We propose a network architecture especially for EEG feature extraction, one that adopts hidden unit clustering with added pooling neuron per cluster. The classification accuracy with features extracted by our proposed method is benchmarked against that with standard power features. Experimental results show that our proposed feature extraction method achieves accuracy ranging from 44% to 59% for three-emotion classification. We also see a 4-20% accuracy improvement over standard band power features.

AB - Spectral band power features are one of the most widely used features in the studies of electroencephalogram (EEG)-based emotion recognition. The power spectral density of EEG signals is partitioned into different bands such as delta, theta, alpha and beta band etc. Though based on neuroscientific findings, the partition of frequency bands is somewhat on an ad-hoc basis, and the definition of frequency ranges of the bands of interest can vary between studies. On the other hand, it is also arguable that one definition of power bands could perform equally well on all subjects. In this paper, we propose to use autoencoder to automatically learn from each subject the salient frequency components from power spectral density estimated as periodogram by Fast Fourier Transform (FFT). We propose a network architecture especially for EEG feature extraction, one that adopts hidden unit clustering with added pooling neuron per cluster. The classification accuracy with features extracted by our proposed method is benchmarked against that with standard power features. Experimental results show that our proposed feature extraction method achieves accuracy ranging from 44% to 59% for three-emotion classification. We also see a 4-20% accuracy improvement over standard band power features.

KW - Autoencoder

KW - Brain-computer-interface (BCI)

KW - Electroencephalogram (EEG)

KW - Emotion classification

KW - Power spectral density

KW - Unsupervised feature extraction

UR - http://www.scopus.com/inward/record.url?scp=85043476858&partnerID=8YFLogxK

U2 - 10.1109/CW.2017.19

DO - 10.1109/CW.2017.19

M3 - Conference contribution

VL - 2017-January

SP - 182

EP - 185

BT - Proceedings - 2017 International Conference on Cyberworlds, CW 2017 - in cooperation with

PB - Institute of Electrical and Electronics Engineers

ER -