Research And Validation Of Sleep Staging Based On Oculoelectric Signals

Today's society is in a stage of rapid development,and more and more workers are under tremendous pressure from work and life.Their sleep are seriously affected by these stress,so it's important to research on sleep.Sleep staging is an important reference indicator in sleep research,providing reliable data for sleep quality assessment and diagnosis.At present,it is clinically concluded that the patient's sleep staging results mainly rely on manual staging by doctors,which is not only inefficient but also susceptible to human factors.The EOG is a very important physiological indicator of sleep staging,so the study of EOG is of great significance.This article extracts the whole night's EOG of the object through self-made acquisition equipment(including eye mask structure,hardware circuit and hardware shell),and the acquisition frequency is 100 Hz.Instead of the conventional disposable electrode patch with Ag Cl,a silver-plated flexible coil is applied to the skin.The method has the advantages of low cost,simple use and good electrical conductivity.The algorithm part is divided into three steps.First,the collected original EEG signal is passed through a Butterworth low-pass filter to filter out part of the high-frequency noise,and then the wavelet threshold de-drying method is used to decompose the low-pass filtered signal and Reconstruction,this step is a pre-processing process,the purpose is to dry the original EOG.Secondly,the multi-scale entropy algorithm is used to extract the eigenvalues of the relatively pure EEG signals obtained in the first step.The specific method is to use the whole night EEG data in units of 30 s and use coarse graining for each unit.Extracting 1?13 different sample entropy.Finally,the extracted feature values are classified by linear discriminant analysis.The sleep stages are divided into five epochs,including Wake,REM,N1 and N2.,N3 epoch.The classification method is a supervised learning of dimensionality reduction technology.The core idea is to project high-dimensional space to low-dimensional and minimize the distance between similar types,and the distance between different classes is the largest to achieve the purpose of classification.The experiment in this paper is combined with the polysomnography in the sleep center of the hospital.The results show that among the 5125 labels,there are 587 correctly classified Wake epoch,the recall rate is 72.83%,and the correct REM epoch is 856.The rate was 71.39%,the number of correctly classified N1 was 903,the recall rate was 72.94%,the correct classification of N2 was 757,the recall rate was 62.98%,the correct classification of N3 was 397,and the recall rate was 58.38%.The system has a recall rate of 68.29% and a Kappa value of 0.60,which is moderately consistent.