Motor Imagery Classification based on Multi-Kernel CNN with the amalgamated Cross Entropy Loss

Jinhyo Shin; Wonzoo Chung

doi:10.1109/BCI53720.2022.9734940

Motor Imagery Classification based on Multi-Kernel CNN with the amalgamated Cross Entropy Loss

Jinhyo Shin, Wonzoo Chung

Department of Computer Science and Engineering

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

4 Citations (Scopus)

Abstract

In this paper, we propose a novel motor imagery (MI) classification method using multi-kernel convolutional neural network (CNN) to optimize the subject dependent kernel size. We designed a multi-kernel CNN consisting of parallel sub-CNNs with different kernel sizes and a weight combiner with the soft-max activation to combine the sub-CNN features. In order to prevent over-fitting problem of the proposed structure, amalgamated cross entropy loss which is defined as the sum of the tentative cross entropy losses of each sub-CNN and the overall cross entropy loss is proposed. The performance of the proposed method is evaluated on BCI Competition IV dataset 2a. The results confirm performance improvement of the proposed method in comparison with existing parallel CNN-based MI classification methods.

Original language	English
Title of host publication	10th International Winter Conference on Brain-Computer Interface, BCI 2022
Publisher	Institute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)	9781665413374
DOIs	https://doi.org/10.1109/BCI53720.2022.9734940
Publication status	Published - 2022
Event	10th International Winter Conference on Brain-Computer Interface, BCI 2022 - Gangwon-do, Korea, Republic of Duration: 2022 Feb 21 → 2022 Feb 23

Publication series

Name	International Winter Conference on Brain-Computer Interface, BCI
Volume	2022-February
ISSN (Print)	2572-7672

Conference

Conference	10th International Winter Conference on Brain-Computer Interface, BCI 2022
Country/Territory	Korea, Republic of
City	Gangwon-do
Period	22/2/21 → 22/2/23

Bibliographical note

Publisher Copyright:
© 2022 IEEE.

Keywords

Brain-Computer Interface (BCI)
Convolutional Neural Networks (CNNs)
Electroencephalogram (EEG)
Motor Imagery (MI)
Multi Kernel

ASJC Scopus subject areas

Artificial Intelligence
Human-Computer Interaction
Signal Processing

Access to Document

10.1109/BCI53720.2022.9734940

Cite this

Shin, J., & Chung, W. (2022). Motor Imagery Classification based on Multi-Kernel CNN with the amalgamated Cross Entropy Loss. In 10th International Winter Conference on Brain-Computer Interface, BCI 2022 (International Winter Conference on Brain-Computer Interface, BCI; Vol. 2022-February). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/BCI53720.2022.9734940

Motor Imagery Classification based on Multi-Kernel CNN with the amalgamated Cross Entropy Loss. / Shin, Jinhyo; Chung, Wonzoo.
10th International Winter Conference on Brain-Computer Interface, BCI 2022. Institute of Electrical and Electronics Engineers Inc., 2022. (International Winter Conference on Brain-Computer Interface, BCI; Vol. 2022-February).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Shin, J & Chung, W 2022, Motor Imagery Classification based on Multi-Kernel CNN with the amalgamated Cross Entropy Loss. in 10th International Winter Conference on Brain-Computer Interface, BCI 2022. International Winter Conference on Brain-Computer Interface, BCI, vol. 2022-February, Institute of Electrical and Electronics Engineers Inc., 10th International Winter Conference on Brain-Computer Interface, BCI 2022, Gangwon-do, Korea, Republic of, 22/2/21. https://doi.org/10.1109/BCI53720.2022.9734940

@inproceedings{32e55dc998f04ca6a209da45a2231f83,

title = "Motor Imagery Classification based on Multi-Kernel CNN with the amalgamated Cross Entropy Loss",

abstract = "In this paper, we propose a novel motor imagery (MI) classification method using multi-kernel convolutional neural network (CNN) to optimize the subject dependent kernel size. We designed a multi-kernel CNN consisting of parallel sub-CNNs with different kernel sizes and a weight combiner with the soft-max activation to combine the sub-CNN features. In order to prevent over-fitting problem of the proposed structure, amalgamated cross entropy loss which is defined as the sum of the tentative cross entropy losses of each sub-CNN and the overall cross entropy loss is proposed. The performance of the proposed method is evaluated on BCI Competition IV dataset 2a. The results confirm performance improvement of the proposed method in comparison with existing parallel CNN-based MI classification methods.",

keywords = "Brain-Computer Interface (BCI), Convolutional Neural Networks (CNNs), Electroencephalogram (EEG), Motor Imagery (MI), Multi Kernel",

author = "Jinhyo Shin and Wonzoo Chung",

note = "Publisher Copyright: {\textcopyright} 2022 IEEE.; 10th International Winter Conference on Brain-Computer Interface, BCI 2022 ; Conference date: 21-02-2022 Through 23-02-2022",

year = "2022",

doi = "10.1109/BCI53720.2022.9734940",

language = "English",

series = "International Winter Conference on Brain-Computer Interface, BCI",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

booktitle = "10th International Winter Conference on Brain-Computer Interface, BCI 2022",

}

TY - GEN

T1 - Motor Imagery Classification based on Multi-Kernel CNN with the amalgamated Cross Entropy Loss

AU - Shin, Jinhyo

AU - Chung, Wonzoo

PY - 2022

Y1 - 2022

N2 - In this paper, we propose a novel motor imagery (MI) classification method using multi-kernel convolutional neural network (CNN) to optimize the subject dependent kernel size. We designed a multi-kernel CNN consisting of parallel sub-CNNs with different kernel sizes and a weight combiner with the soft-max activation to combine the sub-CNN features. In order to prevent over-fitting problem of the proposed structure, amalgamated cross entropy loss which is defined as the sum of the tentative cross entropy losses of each sub-CNN and the overall cross entropy loss is proposed. The performance of the proposed method is evaluated on BCI Competition IV dataset 2a. The results confirm performance improvement of the proposed method in comparison with existing parallel CNN-based MI classification methods.

AB - In this paper, we propose a novel motor imagery (MI) classification method using multi-kernel convolutional neural network (CNN) to optimize the subject dependent kernel size. We designed a multi-kernel CNN consisting of parallel sub-CNNs with different kernel sizes and a weight combiner with the soft-max activation to combine the sub-CNN features. In order to prevent over-fitting problem of the proposed structure, amalgamated cross entropy loss which is defined as the sum of the tentative cross entropy losses of each sub-CNN and the overall cross entropy loss is proposed. The performance of the proposed method is evaluated on BCI Competition IV dataset 2a. The results confirm performance improvement of the proposed method in comparison with existing parallel CNN-based MI classification methods.

KW - Brain-Computer Interface (BCI)

KW - Convolutional Neural Networks (CNNs)

KW - Electroencephalogram (EEG)

KW - Motor Imagery (MI)

KW - Multi Kernel

UR - http://www.scopus.com/inward/record.url?scp=85146201631&partnerID=8YFLogxK

U2 - 10.1109/BCI53720.2022.9734940

DO - 10.1109/BCI53720.2022.9734940

M3 - Conference contribution

AN - SCOPUS:85146201631

T3 - International Winter Conference on Brain-Computer Interface, BCI

BT - 10th International Winter Conference on Brain-Computer Interface, BCI 2022

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 10th International Winter Conference on Brain-Computer Interface, BCI 2022

Y2 - 21 February 2022 through 23 February 2022

ER -

Motor Imagery Classification based on Multi-Kernel CNN with the amalgamated Cross Entropy Loss

Abstract

Publication series

Conference

Bibliographical note

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this