Research On Sleep Characteristics Analysis And Sleep Staging Methods Based On Physiological Signals

Sleep staging is the process of dividing the sleep stage based on the physiological laws during sleep.Accurate staging can provide reliable technical support for sleep quality assessment and related disease diagnosis.The early sleep staging is done manually,subjectivity is strong and the staging efficiency is low.The automatic sleep staging technology that has emerged in recent years,although it largely solves the problem of low work efficiency,it also brings with it contradictions between the number of parameters and the accuracy.This article takes the single-channel EEG and ECG signal from the Physio Bank database as the research object,focusing on the problem of too many parameters and low accuracy in the study of sleep staging technology.The main research content is as follows:(1)Sleep physiological signal feature extraction.In the time domain,frequency domain and nonlinear domain,single-channel EEG features were extracted to obtain 14 characteristic parameters such as the band ratio and sample entropy.At the same time,ECG signals are preprocessed to generate heart rate variability signals,and 9 feature parameters are extracted in multiple analysis domains.(2)Sleep physiology signal feature selection and feature dimensionality reduction.Using adaptive genetic algorithm for feature parameter selection,the accuracy rate is improved by 3% compared to using all feature parameters.Comparing the correlation coefficient method and the basic genetic algorithm,using the parameters selected by the adaptive genetic algorithm can establish a mathematical model with better generalization ability.Performing principal component analysis processing on the selected feature parameters can reduce the feature dimension by half,which provides a possibility for the real-time process of automatic staging of sleep.(3)Classification of sleep physiological signal characteristics.A combined classifier based on genetic algorithm was proposed to optimize the weights of the sub-classifiers on each stage category.Compared with the classification results based on Bayesian voting method and single classifier,the experimental results showed the staging of two combined classifiers.The accuracy rate is higher than that of the four single classifiers,while the combined classifier based on the GA-CMC algorithm is 2.3% more accurate than the Bayesian voting method.The research shows that the GA-CMC automatic sleep staging algorithm based on single channel EEG and ECG signals can use less characteristic parameters and obtain relatively satisfactory classification results.It provides a possibility for real time monitoring of sleep process and has certain application value.