Algorithm Research And Low Power Consumption Design Of Sleep Monitoring System

Sleep is one basic but almost the most important physiological activities for human beings.Moreover,Sleep quality is one of the key criteria to measure their health.As is well-known,sleep hazards and related diseases are prevalent in the crowd,sleep problem hence has arouse social concern.It becomes practical significant to monitor and study sleep in discovering the etiology of the disease,guiding people to maintain a healthy lifestyle and preventing sleep-related chronic diseases,etc.In order to stage sleep and analyze sleep quality of the user,sleep monitoring system usually monitors various physiological signals or other useful signals during sleep.At present,most researches focus on signals selection and features extraction,and use machine learning classification algorithm to solve the most critical sleep staging problem.In practical applications,traditional equipment for monitoring sleep is mainly used for medical purpose and the monitoring procedure often results to much mental distress.Furthermore due to the expensive cost,complex operation and too many sensors are needed,which greatly restrict the sleep monitoring to become popular and to be adopted easily by families or communities for daily demand on health.At present,there are few sleep monitors can satisfy for the family demand or can be popularized as generic products.Most of the similar systems are being only experimented in research stage,and most products in the market has such shortcoming as low accuracy or sometimes not to work propery in recognizing sleep status.The purpose of this study is to research and improve the accuracy of algorithm or model used in sleep staging,and design a home-based,full-day working,real-time,low-power and non-interfering sleep monitoring system.The main work of this paper includes that:proposed a fast hidden layer optimization extreme learning machine,which improves solution accuracy,generalization ability and stability but rarely increases time complexity.and use comprehensive judgment method of classification and prediction to solve the specific issues of sleep stages.This article also studied and researched the whole system of sleep monitoring,designed and implemented the critical algorithm part,and constructed a low-power non-interfering sleep monitoring system.Detail works are specifically as follows:1.Presents a fast and comprehensive hidden layer optimization extreme learning machine algorithm.The proposed algorithm can train output layer network parameters of a number of different hidden layer nodes at each run,avoiding a large number of repeated calculations and guaranteeing the learning effect of the algorithm on the number of hidden layer nodes.Experimental results show that compared with the distributed extreme learning machine,the improved algorithm greatly improves the accuracy of solution,generalization ability and stability,but rarely increases the time complexity.It is very suitable for solving large-scale data classification problems.2.Sleep stage actually contains a recessive timing relationship,so my design uses the method of combining data classification and historical data to improve classification approximation.Experiments have shown that such a design can effectively reduce the impact of uncertainties on the system and increase the reliability of the system.At the same time my design optimize the combination of the two methods,avoid duplicate calculations as much as possible.3.Studied and researched the entire system of sleep monitoring systems,designed and implemented critical algorithms and main procedure of sleep monitoring system,including selection and processing of the monitoring signals,feature extraction and selection of raw data during sleep stages,the quantification of sleep environment scores,and the methods of sleep quality analysis.4.design a home environment based,just in time,full day working,low power consumption sleep monitoring system.