Classification And Recognition Of Body Movement Based On Signals From Micro-Movement Sensitive Mattress Sleep Monitoring System

About one third of a life is spent in sleep,which is an important physiological state.The quality of sleep is quite significant for human health.Related data shows that there are approximate thirty million SAHS patients in our country.In adults,about four percent of male and two percent of female suffer from SAHS.In addition,medicine researches in recent twenty years have proved that,many substantial diseases of human in modern times,such as hypertension,coronary artery disease(CAD), cardiac arrhythmia(CA),diabetes mellitus(DM),cardiac-cerebral vascular accident,and psyche affection,are frequently related to SAS in sleep.Therefore,sleep monitoring technique has been essential to modern medical diagnosis.As one of the main recording parameters in human sleep analysis, body movement is considered to be an important factor for evaluating sleep quality.Frequent movement takes great influence on sleep,because movements,which accompany with the occurrence happening of arousal, can make our sleep quality worse.Body movement during sleep can be used to detect sleep apnea,and muscle convulsion of newborns.Now researches in the unconstrained measurement of body movement are developing rapidly in the world.But an integrity method for recognizing and classifying body movements during sleep with no constraint and low burden has not come out until now.It is known that there exists great relation between the frequency of body movements during sleep and sleep structure.In fact,a lot of information related to human physiological and mental status is contained in body movement.Mining the information deeply can certainly facilitate sleep evaluation into a new stage.This paper begins from this objective, and researches the strategy and realizing method for classifying and recognizing body movement,which is then used in sleep quality evaluation and SAS auxiliary diagnosis.This research is done based on the platform of MSMSMS.The main works of this thesis are as follows:1) Signal preprocessing method is built based on the new generation of MSMSMS with 16 bits A/D.This paper introduces how to design filters after analyzing the fundamental principles of filtering,and considers the problems and solving methods for filter design from the angle of practical application.Then reasonable and effective filters are worked out and used to preprocess the original signals from hardware system.After the preprocessing,several signals reflecting human fundamental physiological parameters human are obtained,and they establish good signal foundation for future research work.2) The classification and recognition of body movement under the condition of unstrained sleep are realized.Aiming at this target,this paper completes feature extraction and Parameter quantification of signals from MSMSMS firstly,and then institutes basic strategy of stratified body movement classification.Methods are searched from two different angles,i.e.,classification with guide and without guide.Then classification of different body movements is explained in detail with window threshold method and machine learning method.3) Comparison experiment is designed to analyze the accuracy rating of body movement categories from the two methods.In the experiment, wireless internet camera records video information of subjects during sleep all night simultaneously,which is used to analyze and compare the signs obtained from window threshold method through artificial cognition. As to the results from machine learning method,artificial data set is also used to cross validation,and error rate or accuracy rate is a fine factor to evaluate learning result.4) Sleep-related breathing events which are automatically distinguished by special software,are modified with the snore signals recorded by microphone simultaneously.First,voiced and unvoiced segments are detected using the method of speech signal processing based on short time energy calculation.Then rules are established and applied to modify the position and length of sleep-related breathing events.This research can realize the accurate detection of breathing events.5) According to the results of body movement classification,this paper completes the compensation of respiratory wave during gross movements,and the correction of breathing events detected by mistake as the result of gross movements.The intervals of gross movements are firstly searched out,and template compensation method is then applied to invalid respiratory wave.Results show that it is effective for correcting mistaken events.This research is done based on the accomplishment of classification and recognition of body movements.On one hand,it reflects significant value of body movements recognition,and on the other hand,it establishes foundation for accurate detection of breathing events in the use of respiratory wave.