Research On Sleep Monitoring Technology With Multi-sensor Fusion

Safeguarding traffic safety is the top priority of railway transportation management,railway department passengers in the work of distraction or fatigue and other conditions often bring tremendous potential safety hazards to driving.In response to this problem,the railway locomotive section has gradually begun to use the fretting sensitive mattress to monitor the sleep of the squad during the forced rest to understand the sleep state of the person.However,the system has problems such as fewer types of biomedical signals collected and lower accuracy.Therefore,this thesis has made some improvements on the basis of the original sleep monitoring system.The main work includes:1.The commonly used multi-sensor fusion methods are analyzed,the tools used in building the soft,ware platform are analyzed,and the principle of signed processing is analyzed;2.Completed the construction of the sleep monitoring platform.On the hardware side,on the basis of the fretting sensitive mattress which only includes the piezoelectric sensor and the acceleration sensor,a finger-clip oximeter for monitoring the blood oxygen saturation signal is added.The related module circuits of the whole sleep monitoring system are analyzed,including the blanket information acquisition module circuit,blood oxygen acquisition module circuit and network transmission module circuit,bluetooth transmission circuit and system transmission protocol are designed;in software,adopt C/S architecture based on MySQL relational database,analysis of sleep physiological data by MATLAB;using C#language programming in Visual Studio,with access to the database to achieve the display,update and storage of data;3.A multi-sensor sleep data fusion model based on D-S evidence theory is established.Based on the learning of the relationship between sleep staging and heart rate,respiration rate,body motion signal and blood oxygen saturation signal,the establishment of sleep knowledge base is completed,the method of multi-resolution analysis based on wavelet transform is used to complete the extraction of sleep information,the extracted sleep information is quantified and analyzed by multi-valued logic,and the corresponding reliability is given according to the degree of validity of the feature.Fusion calculation of all sleep evidences in turn,and the sleep phase with the largest basic probability value is taken as the sleep stage decision result of the period,it is subdivided and corrected according to the general sleep law,that is,the final sleep staging result is obtained;4.Using the relevant data in the MIT-BIH database to verify the model,the obtained sleep staging results were compared with the PSG sleep staging results,and the accuracy was 83.2%;The experimental error is analyzed,and the method to improve the accuracy is explored;Many employees of a locomotive section of the Jinan Railway Bureau were tested for sleep monitoring,and the quality of sleep was judged based on the fusion results.