Research On Sleep Staging Method Based On Ultra-wideband Radar

With the progress and development of society,more and more people enjoy the convenience brought by the times,but also bear a lot of pressure.Therefore,many people begin to have more or less health problems,including sleep problems.According to the survey,a large number of young people are suffering from sleep deprivation and sleep disorders.Corresponding sleep improvement products and sleep medicine have also begun to enter the public's attention.Healthy sleep has become one of the main pursuits of peopleTraditional sleep medicine mainly uses contact sleep monitor to diagnose sleep disorders by wearing electrodes on human body,collecting physiological information into sleep,and then analyzing physiological information and video information by professional doctors.However,this method usually has some shortcomings,such as expensive equipment,complex operation,high requirement for doctors and low diagnostic efficiency,which make it difficult to popularize.And for the first time,sticking a large number of electrodes on the head can make it difficult for people to fall asleep,which also has a great impact on the test results.Therefore,it is of great significance to design a convenient,accurate and suitable sleep monitoring method for family use.In this paper,a sleep staging method based on UWB radar is introduced.A method of extracting body movement information and sign information from radar echo is proposed.A sleep staging method based on BP neural network and a sleep staging method based on support vector machine are proposed.The main contents of this paper are as follows(1)According to the Doppler characteristics of radar,this paper chooses to analyze the frequency domain properties of radar echo and extract the body motion information.Through experimental comparison,the feasibility of extracting body motion information by frequency domain analysis method is verified.Data filters are selected and designed to filter the original radar echoes.For the filtered signal,the short-time Fourier transform method and eigenvalue extraction method are used to extract the sign signal,namely respiratory signal and heart rate signal.After comparing the experimental results,this paper chooses the combination of short-time Fourier transform and eigenvalue extraction to extract the sign information,and verifies the accuracy of the extracted information through experiments(2)Introduced the basic work of polysomnography monitoring system According to the basic basis and rules of current sleep stages,designed sleep data acquisition experiment,collected sleep stages data and radar echo data based on PSG during sleep period.The characteristic parameters of sleep stages are selected from the analysis of the physical and physical information.According to the sleep stages of the PSG,the data are calibrated and the sample data sets are made(3)Introduce how to use BP neural network and support vector machine for sleep staging.The average sleep staging accuracy of BP neural network and SVM is 76.18%and 86.92%respectively.Then the different sleep staging methods are compared through experiments,which proves that the method selected in this paper is feasible,and also reflects the potential of non-contact measurement based on ultra-wideband radar in solving sleep staging problems.Finally,the software designed in this paper can record,analyze and display data in real time,and generate sleep staging reports.