PLSleepNet: A Single Channel EEG Sleep Staging Method Based on Feature Pyramid and Bidirectional LSTM

Sleep staging refers to dividing the sleep process into different stages based on changes in EEG signals, which is of great significance for diagnosing sleep disorders. In previous studies on automatic sleep staging, researchers have achieved good results using multiple channels of EEG for automatic...

Full description

Saved in:
Bibliographic Details
Published in2023 8th International Conference on Intelligent Computing and Signal Processing (ICSP) pp. 189 - 193
Main Authors Wang, Fatong, Gong, Yulin, Lv, Yudan, Liu, Chang, Han, Bo, Li, Tianxing
Format Conference Proceeding
LanguageEnglish
Published IEEE 21.04.2023
Subjects
Bi-LSTM
Brain modeling
Computational modeling
Feature extraction
feature pyramid
Psychology
Signal processing
single channel eeg
Sleep
sleep staging
Time-frequency analysis
Online AccessGet full text
DOI10.1109/ICSP58490.2023.10248895

Cover

More Information
Summary:Sleep staging refers to dividing the sleep process into different stages based on changes in EEG signals, which is of great significance for diagnosing sleep disorders. In previous studies on automatic sleep staging, researchers have achieved good results using multiple channels of EEG for automatic sleep staging. However, when collecting multi-channel EEG signals, it can bring serious psychological and physiological burden to the subjects, which to some extent affects normal sleep and leads to inaccurate sleep monitoring. Therefore, it is necessary to study automatic sleep staging using single channel EEG. However, the limited sleep information contained in single channel EEG poses a certain challenge in extracting effective features and achieving accurate sleep staging. To this end, we propose a neural network model called PLSleepNet, which utilizes feature pyramids and attention mechanism based Bi-LSTM for automatic sleep staging of single channel EEG. The feature pyramid can fully extract features at different time and frequency scales, maximizing the mining of sleep information in single channel EEG. Based on attention mechanism, Bi-LSTM can capture temporal features and contextual information, and give different weights according to the importance of different sleep stages, thereby improving the accuracy of sleep staging. To verify the effectiveness of PLSleepNet, we conducted experiments on the open dataset Sleep-EDF Database Expanded and compared it with five other staging models. The experimental results show that PLSleepNet outperforms other methods in overall accuracy, MF1 score, and Kappa coefficient, reaching 84.5%, 78.0%, and 0.786.
DOI:10.1109/ICSP58490.2023.10248895