Human Sleep Monitoring Methods And Sleep Quality Analysis

Good sleep is a necessary condition for recovery,mental and physical health.The quality of sleep is closely related to the physical and mental health of human beings.People have gradually realized the importance of sleep health,and know that the distribution of sleep stages plays a crucial role in the quality of sleep.So people design convenient,popular wearable devices to monitor sleep.The wearable device has the problems of anti-interference and availability,low accuracy,etc.The data produced by the wearable device is characterized by high noise and low quality.In order to get better sleep analysis results on this type of low-quality data,design and implement a simple and easy-to-use sleep pillow to monitor sleep.In the three directions of sleep data preprocessing,feature design and selection and classification method,the purpose of higher sleep classification accuracy is achieved.The main research contents are as follows:1.Sleeping pillow to obtain the original data have the characteristics of aperiodic,finiteness,discrete,to obtain the implied the respiration signal,the discrete Fourier transform and the basic principle of discrete wavelet transform is put forward based on discrete Fourier the respiration signal reconstruction method and based on the discrete wavelet transform of the respiration signal reconstruction method.To describe the respiration signal details on time domain and frequency domain information,combined with the advantages of FFT can accurately obtain the frequency spectrum component and DWT can accurately in time domain to obtain the advantages of each frequency band signal,puts forward the respiratory signal acquisition method based on FFT + DWT.2.The design of features and selection are very important to the classification effect.In this paper,body dynamic extraction algorithm and respiratory peak detection algorithm are proposed to obtain the body movement information and respiratory information.In the body dynamic signal,the design characteristic describes the momentum of the body,the duration of body movement and the dynamic distribution of the body.In order to capture the effect of respiratory change on sleep stage,the relationship between respiration and sleep state was described in the time domain and frequency domain.The four feature selection methods of f-score,FCBF,mRMR and LLDR are implemented to screen the features and eliminate redundant features,and verify the characteristics of the design.3.Single classifier for data classification and cannot achieve good results,this paper described to the largest extent with poor data quality physiological signals and the link between the sleep stage,in terms of sleep stages division,according to the physiological signals in each stage of sleep,the difference of the design of binary decision tree bayesian classifier combination and selection method.Improved classification accuracy through decision making and classifier integration.