基于CNN-LSTM网络的睡眠分期研究

Sleep classification based on CNN-LSTM network

睡眠分期是睡眠数据分析的基础,用自动标定方法来替代人工标定方法可以提高效率,结果也更为客观。不管是人工手动标定还是自动标定都是基于多导睡眠图(Polysomnography,PSG)。采用长短时记忆模型(LSTM-RNN)及长短时记忆模型与卷积神经网络(CNN)相结合的模型(CNN-LSTM)对三个通道信号(EEG、EOG、EMG)的组合进行自动睡眠分期研究。通过对9个受试数据进行分析,LSTM-RNN和CNN-LSTM分别达到了81.9%和83.1%的分类准确率。相对于LSTM-RNN模型,结合卷积神经网络的CNN-LSTM模型获得的分期准确率更高,平均准确率提高了1.2%。

Sleep classification is the basic step of sleep data analysis. The automatic classification method can improve the efficiency and obtain more objective result compared with manual calssification method. Automatic or manual classification of sleep stage are based on the Polysomnography(PSG). In this paper,we employed the long short-term memory recurrent neural network(LSTM-RNN) and the long and short memory model combined with convolution neural network(LSTM-CNN) to perform automatic sleep classification with multichannel sleep data(EEG,EOG and EMG). In 9 subject's sleep data,the classification accuracy were 81. 9% and 83. 1% by using LSTM-RNN and LSTM_CNN respectively. The CNN-LSTM method increased 1. 2% accuracy than LSTM-RNN.