Research On Physical Fatigue Prediction Based On ECG Signals

In recent years,with the continuous enhancement of living standards,the work intensity of people in daily work have increased,which may worsen the trend of hypertension,depression,and cardiovascular and cerebrovascular diseases at a younger age.The concept of exercise for health has taken root by enhancing people’s physical fitness to reduce the impact of diseases.However,there is still a lack of awareness about scientific and reasonable physical exercise.For example,certain inappropriate physical activities are likely to cause physical fatigue,which to a certain extent can damage human body functions and thus lead to a decline in work performance.If a wearable device can monitor a physiological indicator in real time to predict the level of physical fatigue,and give people a timely reminder to adjust the intensity of work reasonably,then it can effectively avoid adverse effects and maintain human health and safety.The research on physical fatigue is still at an initial stage,and the paper conducted a study on wearable ECG based on the problem of lack of proper means for timely,convenient,and effective physical fatigue monitoring and assessment,and the results have important practical applications.In each cycle of heartbeat,the ECG signal,as a bioelectric signal reflecting the change of potential inside and outside the human cardiomyocyte membrane,is suitable for mobile wearable personal continuous detection scenarios and has been widely used in the fields of fatigue monitoring,cardiac arrhythmia,and rehabilitation training.The thesis is based on the prediction study of fatigue state and aims to explore effective physiological indicators and methods for predicting physical fatigue by selecting some college athletes and analyzing and evaluating physical fatigue using ECG signals,combining subjective fatigue perception and objective cognitive performance.The specific research content and results of the paper include the following.(1)A"training-cognitive task"physical fatigue experimental data collection protocol was designed using a portable electrocardiographic acquisition device(Shimmer3)to collect physiological signals from subjects during physical training.To facilitate the data collection,the subjects wore the device during the one-hour rope skipping training and were required to complete a Samn-Perelli scale-based self-fatigue assessment and an n-back cognitive task before and after the training,which aimed to reflect the subjects’physical fatigue level from both subjective and objective levels and provided data support for the subsequent physical fatigue analysis.(2)Statistical analysis of ECG indices was performed.To address the problem that the traditional R-peak detection method of ECG signal may not be accurate due to the phenomenon that there are significant differences in the heart rate of subjects before and after training.The paper adopts an improved adaptive sliding time window-based algorithm to detect R peaks and extract RR intervals,and then analyzes the characteristics of heart rate variability(HRV)of participants in the time domain,frequency domain,and nonlinear dynamics of ECG signals,and Pearson performs correlation coefficient analysis and paired t-test on the extracted indexes,from which statistically significant ECG indexes are selected.(3)Autonomic mechanisms of physical fatigue were analyzed.In order to confirm whether subjects reached a state of physical fatigue before and after physical training,some existing literature on physical fatigue analysis was consulted,and common methods were used to account for physical fatigue through specific scores on subjective fatigue questionnaires or significant differences in statistical analyses of pre-and post-fatigue.Therefore,the paper used a one-way ANOVA to analyze the significance of subjects’self-fatigue assessment and performance on n-back tasks before and after physical training.Based on the results of significant differences to account for the significant effects of training on subjects’subjective fatigue and objective cognitive performance,the autonomic mechanisms of physical fatigue were analyzed to provide a theoretical basis for predictive studies of physical fatigue.(4)Machine learning prediction models were constructed.Particle swarm optimization-based support vector regression(PSO-SVR)and extreme learning machine(ELM)regression models for predicting subjective and objective physical fatigue were constructed,and regression evaluation metrics such as mean absolute percentage error(MAPE),root mean square error(RMSE),and coefficient of determination(R~2)were used to validate the ability of different models for predicting physical fatigue,while the prediction performance of machine learning models based on Gaussian process regression(GPR),random forest(RF),k-nearest neighbor(KNN),and adaptive boosting(Ada Boost)were also compared.The results of the paper showed that PSO-SVR performed better in predicting subjective physical fatigue with a MAPE of 11.53%and RMSE of 0.2842,or R~2 of89.26%,followed by Ada Boost,which also showed good prediction results with a MAPE of 16.95%and RMSE of 0.5322,or R~2 of 63.29%.The ELM,on the other hand,achieved competitive results in objective physical fatigue prediction,with MAPEs of 22.42%and 19.77%for the mean response times(ORT and TRT)of 1-back and 2-back in objective cognitive levels,respectively.The research in the thesis provides a reliable experimental basis and useful theoretical analysis conclusions for the effective identification of physical fatigue.