Multimodal Sleep Staging And Sleep Disease Diagnosis Based On Physiological Signals

In the context of rapid social and economic development,personal stress continues to increase,and the prevalence of sleep diseases continues to increase.Sleep monitoring and home observation of sleep disease diagnosis have become serious problems that need to be resolved in sleep research.Therefore,the use of a single-channel physiological signal with a simple extraction process for research in sleep-related fields has practical application value.This paper presents a simple and effective method for multi-class automatic sleep staging and sleep disease diagnosis based on Photoplethysmography(PPG).First,the photoplethysmographic signals of 108 subjects in the CAP sleep database were preprocessed and the multi-modal physiological signals were isolated out from the preprocesed signals;then,the features of multi-modal signals were extracted from the time domain,frequency domain,and non-linear methods.Signal feature extraction.Then,Light Gradient Boosting Machine(Light GBM)was used to classify multi-type sleep staging.Finally,in the part of sleep disease diagnosis,SVM classifier was applied to sleep disease diagnosis after data preprocessing and feature extraction from time domain and using nonlinear method,based on the four-classification results of the subjects' sleep monitoring throughout the night.The innovation points of this study are listed as follows:1.In the process of PPG signal acquisition,signal interference appears caused by body motion,dark current,ambient light,etc..This study uses the biorthogonal spline wavelet method to filter baseline drift and power frequency interference in the signal.2.Multi-modal signals such as heart rate signals and heart rate variability signals are extracted from the PPG signal,and the multi-modal signals are extracted from the time domain,frequency domain,and nonlinearity domain.3.The accuracy of tri-classification,four-classification and five-classification sleep staging was 88%,83% and 78%,respectively.and the corresponding kappa coefficients are 0.79,0.76 and 0.70.Moreover,the method of this study is also applicable for subjects with sleep disorders.4.In this study,the results of the above four categories of subjects' sleep stages are used as data to realize the distinction between healthy subjects and subjects with sleep diseases in the CAP sleep database.