Construction Of Sports Fatigue Detection System Based On Multi Physiological Information Fusion

Since the beginning of the 21 st century,as people’s attention to health issues has increased significantly,exercise and fitness have become a trend in life.However,it is worth being vigilant that there are endless cases of excessive exercise causing harm to the body.Therefore,real-time adjustment of exercise intensity through exercise fatigue status plays an important role in effectively preventing potential safety hazards.In this thesis,a set of exercise fatigue detection system based on the fusion of multiple physiological information is designed and produced.It uses multiple physiological signal acquisition equipment to collect data while walking,and sends it to the PC through the master-slave Bluetooth module HC-05,and the physiological information,walking steps and energy consumption of the subjects are displayed on the PC in real time.The relationship between the physiological signals and the fatigue of the movement was analyzed from the point of view of statistics and machine learning,and the fusion mode of physiological signals which could reflect the fatigue level was obtained.This topic draws on the method of labeling fatigue levels in the research of "Treadmill-based Cardiopulmonary Endurance Test" conducted by Xu Hang’s team at the University of Science and Technology of China,and divides the fatigue levels into five levels.During the experiment,six physiological signals corresponding to each fatigue level were collected,including heart rate,blood oxygen saturation,back skin temperature,and waist X,Y,and Z acceleration signals that can reflect changes in gait.Taking physiological signal as independent variable and fatigue level as dependent variable for statistical analysis,the fusion mode of physiological signal which can best reflect fatigue degree is obtained,and a linear expression of "fatigue level physiological signal" is fitted.At the same time,a variety of commonly used algorithms in supervised learning are used to establish a fatigue recognition model,and the physiological signal fusion method with the highest accuracy of fatigue discrimination is obtained.In the actual test process,it was found that the linear expressions fitted to five signals of heart rate,blood oxygen saturation,temperature,waist Y-axis acceleration,and waist Z-axis acceleration can most effectively determine the fatigue level.The test set of the fatigue detection model established by random forest,gradient boosting tree,and support vector machine has an accuracy rate of more than 95%.Among them,the gradient boosting tree has the best effect on sports fatigue recognition based on the fusion of heart rate,blood oxygen saturation,temperature,waist Y-axis acceleration,and waist Z-axis acceleration,with an accuracy rate of 96.00%.