Design Of Stroke Monitoring And Recognition System Based On Wearable Sensor Network

Swimming can improve the function of human respiratory system and cardiovascular system.The monitoring and recognition of swimming is conducive to monitoring and controlling the training process of athletes,as well as sports injury recovery and disease rehabilitation treatment.At present,wearable system based on inertial sensor is widely used for swimming posture monitoring and recognition.However,the existing research mainly focuses on the wrist,back and so on,and there is less research on the position of legs.Moreover,the experimenters fix the sensor on the human body by binding and pasting.The sensor is easy to fall off during movement,and the connecting wire is exposed,which affects the use.In view of the above problems,this paper designs and makes a swimming posture monitoring and recognition system based on wearable sensor network.The main work includes:1.Starting with the theoretical knowledge of swimming,determine the model and location of the sensor,and develop an interactive interface based on Lab VIEW,which can quickly and reliably display the real-time changes of knee and hip angle information and leg and waist angle information of four competitive strokes（freestyle,backstroke,breaststroke and butterfly）,and build a stroke lower limb motion monitoring system based on inertial sensor.2.The wearable device is designed to waterproof the electronic components.The integrated fabric sensor network is woven according to the double arrangement sensor network model,and the connected wires will not be exposed.The detachable design of the sensor pants is realized by using elastic Velcro,which can effectively solve the problem of wire joint damage that may occur in the wearing process,and broaden the use scene of the acquisition system.3.The wearable acquisition system is used for data acquisition,and the action data of four strokes of 10 subjects are collected.The similarities and differences of the data of the same type of stroke and different strokes are compared,and the stroke recognition is preliminarily realized.4.The stroke recognition function is realized through modules such as data preprocessing,information combination,classification and recognition and online recognition.Different information combinations and classifiers are trained respectively,and evaluated through classification evaluation indicators.It is concluded that the DT classification model based on the combination of all motion information has the best recognition effect,and the recognition accuracy is 99.51%;The combination of sensor node Angle information has the best recognition effect for the four swimming strokes,and the average recognition accuracy is 94.52%.At the same time,the action of different stages of breaststroke was extracted for classification recognition,and the available range of recognition model was broadened.Finally,the trained recognition model is used for on-line recognition,which verifies the feasibility and effectiveness of stroke monitoring and off-line recognition system.