Research On Sleep Monitoring Methods Based On Sound And ECG Signals

Sleep plays an important role in the health of the human body.Lack of sleep can lead to fatigue and lack of concentration during the day.Besides sleep,the quality of sleep is also an important factor in maintaining health.Clinical studies show that sleep is associated with many serious diseases,including diabetes,obesity and depression.In recent years,with the popularization of high quality sensors,there are some researches on the low cost sleep assessment system.The sleep analysis system based on the voice signal of mobile phone has become a cheap and effective alternative to high polysomnography.Obstructive sleep apnea(OSA)is characterized by periodic respiratory arrest(pause)and periodic hypoventilation.This is caused by the collapse of all or part of the airway during sleep.When there is no air entering the pulmonary artery,the blood oxygen level drops and the level of carbon dioxide increases,which leads to deoxidization,followed by the excitement of the sympathetic nervous system,which eventually causes the awakening.In the long run,it may cause sleep problems and cardiovascular diseases.In this paper,we use sound signals and ECG signals to analyze the sleep status and sleep quality of the users,and screen the OSA events.The specific work is divided into the following three aspects:(1)Sleep analysis based on sound signals.In this paper,the sound data is collected by the smart phone,the noise is modeled on the original data,the features are extracted after the denoising,and then the multi classification support vector machine is used to classify various sleep events,and the quality and disturbance factors of the sleep are analyzed.(2)Research on apnea diagnosis method based on sound signal.In this paper,the sound data is collected by the smart phone.The V-Box algorithm is used to filter the unrelated data frames,and the characteristics of the filtered data are extracted.The fuzzy decision tree is used to classify the snoring events and non-snoring events,and then the characteristics of the snore data frames of different users are extracted,and the support vector machine is finally used for the classification of OSA patients and non OSA patients.(3)Research on apnea detection method based on ECG signals.In this paper,the OSA event is detected by ECG signal,the time domain,frequency domain and nonlinear characteristics of RR intervals are extracted,and the features are screened by the bidirectional search(BDS)algorithm and the minimum redundancy maximum correlation(mRMR)method.In this paper,support vector machines(SVM),random forests(RF)and AdaBoost are used in classification,and the comparison is made from the classification effect.