| Nowadays,with the improvement of human living standards,more and more chronic diseases caused by the imbalance of energy metabolism,including obesity,diabetes,hyperlipidemia and cardiovascular diseases,have become the focus of attention all over the world.More and more people are aware of the importance of exercise to health and maintaining regular physical exercise to keep healthy.Active health management includes the scientific control of dietary energy intake and physical activity energy expenditure,which provides an effective way for the prevention and rehabilitation of chronic diseases.As the human body is a complex time-varying nonlinear system,the energy expenditure during physical activity can be affected by many factors,including activity intensity,individual physiological and psychological state,environmental factors and anthropometric data,making real-time and accurate energy expenditure estimation a challenging research.However,proper exercise is beneficial to health,while excessive exercise may lead to physical injury.Excessive exercise is defined as a relative term,which means that one or more rounds of exercise may be excessive for a specific individual based on a variety of factors such as health status and genetics.Physical fatigue is a common physiological phenomenon during physical activity and a direct reflection of the degree of exercise.Accurate detection and evaluation of physical fatigue levels can effectively prevent excessive exercise and further reduce physical injury caused by excessive exercise.However,as a subjective feeling of individuals,physical fatigue is difficult to be objectively evaluated.Due to the existence of the above problems,this paper designs experiments to collect the ECG signals and inertial sensor signals in the process of human running,and proposes a regression model for accurate energy expenditure calculation and the classification model of physical fatigue state by using the relevant methods of deep learning and machine learning.Specifically,this paper mainly completes the following two works:(1)A deep multi-branch two-stage regression network model is proposed to effectively integrate various relevant information,including motion information,physiological characteristics and anthropometric characteristics,which significantly improves the accuracy of energy expenditure estimation.The proposed deep multibranch two-stage regression network mainly consists of two modules: A multi-branch convolution neural network module,which extracts multi-scale context features from ECG signals and inertial sensor signals;and a two-stage regression module,which uses the extracted features including physiological and motion information and anthropometric features to calculate energy expenditure.The experimental results show that compared with the previous work,the proposed method is more accurate and the average root mean square error is reduced by 22.8%.The EE estimation accuracy is improved by the proposed DMTRN model with a well-designed network structure and novel input signal ECG.This study verifies that ECG signal can calculate energy expenditure more effectively than heart rate and casts light on EE estimation using the deep learning method.(2)A classification method of physical fatigue based on HRV is proposed.Firstly,a total of 24 HRV features are calculated from ECG signals.Then a feature selection method is proposed and implemented to select the 11 best features.Finally,four machine learning algorithms are trained to classify fatigue using the selected features.The experimental results show that the model trained with the selected 11 features can classify three levels of physical fatigue with high accuracy,and the highest accuracy is85.46%.More importantly,it is also verified that the feature selection method proposed in this paper is of great significance to improve the performance of physical fatigue classification.These selected features can provide important information about the identification of physical fatigue. |
PDF Full Text Request