Analysis On Statistical Feature Of Motion Patterns And Quantitative Study Of Sports Fatigue Based On Physiological Signals

Motion pattern recognition and sports fatigue detection are two very important research tasks in sports health.However,the existing research work has not been able to implement the short-time,high-precision motion pattern recognition and fatigue detection algorithm on wearable devices.In addition,there is a lack of evaluation methods and indicators in the field of fatigue quantification and sports crowd analysis.Physiological signals objectively reflect the motion state of human body and can be conveniently acquired.With good learning and classification ability and strong interpretability,machine learning algorithm is suitable to be implemented in sports health and related research.Therefore,based on short-term multi-dimensional physiological signals,this paper applies a variety of signal processing methods to extract relevant features and study their statistical properties;and use machine learning algorithms to identify the motion pattern and detect fatigue,and builds the fatigue index system and fatigue index,and finally conducts in-depth research and analysis on sports crowds.In this paper,the Bioharness device is used to obtain sports signal,based on which the statistical features of motion patterns are extracted.The XQRS algorithm is improved to locate the R wave of ECG signal.We conduct in-depth statistical analysis on the motion pattern features and apply machine learning algorithms to establish the classification models which realize the classification of the four typical motion patterns of stationary,walking,running and vertical jumping: the accuracy of the XGBoost model based on signal of 1s and 5s reaches 96.77% and 99.05% respectively.In addition,it found that the motion pattern recognition model has a certain "memory" and can identify the past motion pattern accurately.For sports fatigue detection,this paper shortens the time of signals to 30 seconds and significantly improves the detection accuracy.In feature extraction,besides calculating various time-frequency features,the improved XQRS algorithm is combined with Neurokit algorithm to locate the PQRST complex of the ECG signal,and the empirical mode decomposition and complex network analysis methods are also introduced.In this paper,a statistical analysis of the sports fatigue features is carried out,and four machine learning algorithms are applied to establish fatigue detection models with the embedding method.Among them,the XGBoost algorithm model has the best performance,with an accuracy of 99.46% and an F1 score of 99.29%,which basically meets the needs of practical applications.Statistical analysis and model results show that the heart rate,respiratory rate and respiratory depth signals and related features play an important role in fatigue detectionBase on statistical analysis and feature selection of fatigue detection model,a feature system reflecting the fatigue degree of sports is constructed,and an objective fatigue index is constructed by combining classification decision functions of logistic regression and linear kernel SVM.Statistical results prove that the fatigue index can effectively reflect the sports fatigue state and has high application value.When exercising at a certain intensity,the changes of physiological signals and sports fatigue index have different performances in different sports crowds.Stand on this,this paper constructs a logistic regression model for distinguishing sports crowds,and the classification accuracy of which is 94.36%.This work can be used to monitor changes in human body fitness and state,and has a good application prospect.