Method Of Automatic Sleep Staging Based On Portable Sleep Monitoring System

Nearly one-third of a human being’s life is spent in sleep.The quality of sleep is closely related to the working life of a human being,as well as his physical and mental health.In proceeding of the sleep,sleep status will experience several relatively stable stages,clinically known as sleep classification.And sleep electroencephalography(EEG)is the most basic and the most effective way to research sleep classification.With public’s growing attention to sleep issues,portable sleep staging monitoring equipment will be the developmental direction of future sleep monitoring.Due to application environment and user experience,portable sleep monitoring equipment usually use single channel to collect EEG signals.And this thesis mainly conducts an in-depth study on the sleep automatic classification method of portable sleep monitoring equipment based on single-channel sleep EEG signals.Sleep automatic classification is an efficient technology of pattern recognition,and is a key issue in sleep research.This thesis summarizes the common methods of sleep automatic classification by combing its basic theories,and proposes a multi-parametric feature extraction method based on EEG normalization.This method first uses the Stable Wavelet Transform(SWT)threshold denoising to remove the noise in the original EEG signal,and then normalizes the denoised signal as the basis for subsequent research.Secondly,uses Complete Ensemble Empirical Mode Decomposition with Adaptive Noise(CEEMDAN)to analyze normalized signal,obtain its intrinsic mode functions in multiple time scale.Then base on the normalized signal and IMF components to extract the multi-parameter of EEG signal.Finally,Bootstrap Aggregating(Bagging)was performed as classifier to complete the automatic staging of sleep.At the same time,a sleep automatic staging prototype system is designed and implemented based on proposed method.The system mainly includes a file processing module,a signal denoising module,a signal analysis module,a classification obtaining module,and a sleep automatic staging module,providing support to the system operation.In the end,this thesis uses the Sleep-EDF database to evaluate the prototype system,and analyzes the classification and evaluation results.Through the evaluation,we can come to a conclusion that the method proposed in this thesis can be used toachieve highly accurate automatic sleep classification.And the prototype system implemented can demonstrate the automatic staging process,providing technical supports for more mature portable automatic classification intelligent system.