EEG-based user identification system using 1D-convolutional long short-term memory neural networks

•Achieve feature extraction from both spatial and temporal domains of EEG signals.•The proposed network has accuracy up to 99.58% when using 16 EEG channels.•Reduce number of channels of the EEG systems while maintaining high performance. Electroencephalographic (EEG) signals have been widely used i...

Full description

Saved in:
Bibliographic Details
Published in:Expert systems with applications 2019-07, Vol.125, p.259-267
Main Authors: Sun, Yingnan, Lo, Frank P.-W., Lo, Benny
Format: Article
Language:English
Subjects:
1D-Convolutional LSTM
Biometrics
Electroencephalograms (EEG)
Electroencephalography
Fingerprinting
Identification methods
Identification systems
Internet of Things
Neural networks
Short term
User identification
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:•Achieve feature extraction from both spatial and temporal domains of EEG signals.•The proposed network has accuracy up to 99.58% when using 16 EEG channels.•Reduce number of channels of the EEG systems while maintaining high performance. Electroencephalographic (EEG) signals have been widely used in medical applications, yet the use of EEG signals as user identification systems for healthcare and Internet of Things (IoT) systems has only gained interests in the last few years. The advantages of EEG-based user identification systems lie in its dynamic property and uniqueness among different individuals. However, it is for this reason that manually designed features are not always adapted to the needs. Therefore, a novel approach based on 1D Convolutional Long Short-term Memory Neural Network (1D-Convolutional LSTM) for EEG-based user identification system is proposed in this paper. The performance of the proposed approach was validated with a public database consists of EEG data of 109 subjects. The experimental results showed that the proposed network has a very high averaged accuracy of 99.58%, when using only 16 channels of EEG signals, which outperforms the state-of-the-art EEG-based user identification methods. The combined use of CNNs and LSTMs in the proposed 1D-Convolutional LSTM can greatly improve the accuracy of user identification systems by utilizing the spatiotemporal features of the EEG signals with LSTM, and lowering cost of the systems by reducing the number of EEG electrodes used in the systems.
ISSN:0957-4174
1873-6793
DOI:10.1016/j.eswa.2019.01.080