基于深度卷积神经网络和循环神经网络的睡眠分期模型

Sleep Staging Model Based on Deep Convolution Neural Network and Recurrent Neural Network

针对传统机器学习模型过于依赖特征工程、多导睡眠图(Polysomnography,PSG)数据获取难度大等问题,提出一种基于深度卷积神经网络(ConvolutionalNeuralNetworks,CNN)和循环神经网络(RecurrentNeural Network,RNN)的自动睡眠分期模型。该模型不需要烦琐的特征提取过程,仅使用单通道脑电信号即可在较高水准下完成自动睡眠分期,在公开数据集Sleep-EDF的Fpz-CZ通道脑电数据上实现了85.2%的分类准确率。

Aiming at the problems of traditional machine learning models that rely too much on feature engineering and the difficulty of Polysomnography(PSG)data acquisition,this paper proposes an automatic sleep staging model based on deep Convolutional Neural Networks(CNN)and Recurrent Neural Network(RNN).The model does not need complicated feature extraction process.It can complete automatic sleep staging at a high level by using only single channel EEG signals.It achieves 85.2%classification accuracy on the Fpz-CZ channel EEG data of the open data set Sleep-EDF.