Research On The System Of Wearable Health Monitoring For Outdoor Sports

With the rapid development of economy and culture,there are more and more people plunge into outdoor sports and the related industries.Since our country’s outdoor sports industry is still in its infancy,there are many problems in the professional knowledge,skill reserve,outdoor equipment of participants and logistics rescue support and so on,resulting in a high incidence of outdoor sports accidents,and the safety of participants is threatened.Based on multi-source sensor data fusion technology,this thesis designs and implements a wearable health monitoring system to fit outdoor environment,which can monitor the movement,environment and physiological status of participants in real time,automatically identify abnormal status,and promptly sends out an emergency signal after an acidents.The main research work and contributions can be divided into two parts:design of Real-time Physiological Monitoring System based on multi-source sensors and Anomaly Identification Method based on CNN-LSTM fusioned expert knowledge.In the research of the real-time physiological monitoring system,the modular method is used to design the hardware and software.In terms of hardware,the system is composed of several modules such as local play,main control,communication,power management,edge calculation,and four acquisition modules for motion,environmental,physiological and voice signals.In terms of software,the modularized multi-sensor drivers and a real-time monitoring tool for data acquisition,transmission,processing,application and storage are designed and implemented based on the embedded system.Finally,the real-time,long-term and stable monitoring of motion indicators such as acceleration,angular velocity and direction angle,environmental indicators such as temperature and humidity,atmospheric pressure,light intensity and GPS position,and physiological indicators such as BT,PPG,HR,BP and SPO₂ were completed.The system can help participants understand their own situation quickly to facilitate action planning;It helps search and rescue medical staff to keep abreast of the participants,locate the wounded after an accident quickly,and obtain information about the injury and their surroundings,which is greatly beneficial to search and rescue,medical and other operations.This research can provide technical support and safety guarantee for outdoor sports participants,and has certain application value in the field of personal health monitoring.In the research of anomaly recognition algorithm,we have explored the discriminant method which combines multi-source sensor signals with expert experience,and proposed a two-stages scheme of anomaly recognition problems.In the first stage,the discriminant rules containing expert experience are used to complete the preliminary discrimination of anomalies.In the study of expert knowledge base,we have explored the representation and application of expert experience in wearable systems,designed and implemented a construction scheme for expert knowledge base of anomaly discrimination problems based on Q&A mode.In the second stage,the CNN-LSTM network is used to process multi-source sensor signals,which making full use of the feature extraction capabilities of Convolutional Neural Networks and the time-series expression characteristics of Long Short-Term Memory networks,which greatly reduces the underreporting rate of abnormal states.In the application of the algorithm in wearable devices,We have designed and implemented a hot-update algorithm for the model,which implements the hot load of the model without pausing the service,ensuring the long-term and continuity of monitoring task.which realized the eager loading of the model without suspending the service,and ensured the long-term and continuous monitoring.Finally,the effectiveness of the system and algorithm is verified through human experiments and simulation experiments.