一种基于双模态的睡眠分期研究

打开文本图片集

中图分类号：TP399 文献标志码：A 文章编号：1671-6841（2025）03-0081-07

DOI： 10. 13705/j. issn. 1671-6841. 2023257

A Study on Sleep Staging Based on Bimodal Analysis

WANG Yaqun1，23， YANG Qing1，2.3， WEN Dou123，WANG Ying1，23，WANG Xiangyu 1，2，3 （1. Hubei Provincial Key Laboratory of Artificial Inteligence and Smart Learning，Central China Normal University， Wuhan ，China; 2. School of Computer，Central China Normal University， Wuhan 43O079，China; 3. National Language Resources Monitoring & Research Center for Network Media，Wuhan ，China）

Abstract： The existing research generally focused on a single signal， ignoring the sleep information provided by other model signals in a specific sleep stage.The loss of important information，as the network deepened when extracting sleep signals，could reduce the clasification ability of the model in view of the problems，a deep neural network model based on electroencephalogram（EEG）and electrooculogram （EOG）was designed to sleep stage in an end-to-end manner， which was called MCNN LSTMs model. The features extracted from EEG and EOG signals were fused by a two-layer long short term memory （LSTM）neural network after multi-scale convolutional neural network，and then input into the classifier for sleep staging. The performance of the proposed method on sleep staging was evaluated on a public sleep EDF dataset. Experiments showed that when two channels （EEG-EOG） were used，the clasification accuracy reached 92.60% on Sleep-EDF-20 dataset and 91.10% on Sleep-EDF-78 dataset，which was better than single channel signal and comparison methods. The effctiveness of multiple signals for sleep staging was verified，and an important idea for the study of sleep staging was provided.

Key words： sleep stage ; multimodal; convolutional neural network ；multi-layer network ； LSTM

0 引言

睡眠几乎占我们生命活动的三分之一，良好的睡眠对于维持一个人的身心健康至关重要。（剩余12153字）