A Research Of Real-time Sleep Staging Algorithm And The Development Of Application System

An increasing number of people are suffering from disturbed sleep,evaluating sleep quality to improve the people's sleep conditions has become a major issue.Accurate sleep staging is the basis for assessing sleep quality and diagnosing sleep-related diseases objectively,and the classical automatic staging methods are basically to analyze Electroencephalogram(EEG)signals.However,the operation of the EEG signals is complex,the cost is high,and the placement of the electrodes will interfere with the normal sleep,it cannot meet the needs of family sleep monitoring occasions.It is proposed to use the physiological parameters those are more conveniently detected to automatically analyze the sleep process,which has more practical value compared with the traditional EEG,the use of piezoelectric sensing mattress can achieve multiple sleep parameters in the longtime synchronization test almost without any interference during sleep,so it can monitor people's real sleep conditions at home and has a good application prospects.The mattress collects ballistocardiogram(BCG)signals,including heart rate,respiration rate and body movement information,but the accuracy of automatic sleep staging of the BCG signals is not high at present.The purpose of this study is to establish a more accurate and reliable multi-parameter automatic staging algorithm based on the BCG signals and then apply it to real-time monitoring of sleep in the implemented mattress real-time sleep monitoring system.This paper used the heart rate,respiratory and body movement signals calculated from the BCG signals for automatic sleep staging,which is divided into four stages: awake,light sleep,deep sleep and rapid eye movement sleep.Because the heart-rate waveforms in different sleep stages are more difficult to distinguish,the features of heart rate variability(HRV)were extracted by time-variant autoregressive model,further,the features were automatically classified based on the hidden Markov model.It is found that the combination of the phase of the pole in high frequency band and the total power obtained the best performances of the sleep staging,and combined the respiratory rate and body movement information for the sleep staging correction,which could effectively improve the accuracy based on HRV.Using the data in the MIT-BIH database to test,and compared the results of the algorithm in the paper and the expert's staging to verify the accuracy of the proposed multi-parameter sleep staging algorithm,which has a recognition rate with 70.13% and is computationally fast,therefore can be used for realtime monitoring.This paper designed and implemented a mattress real-time sleep monitoring system,which employed the established automatic sleep staging algorithm and can be used in the home environment to monitor people's sleep in real time.The system is based on the piezoelectric sensing mattress,and the real-time heart rate,respiration rate and body movement data separated from the BCG signals are uploaded to the server for storage,then the automatic sleep staging is completed.Finally,the real-time sleep information,sleep quality analysis,advices for improving it and other services are displayed to users through the smartphone application,as an attempt to help people increase the quality of sleep.