The Identification On Basic Human Motion Based On Multiple Accelerometers Of Strap Distribution

Purpose In the prevention and treatment of obesity-related chronic diseases such as high blood sugar,high blood pressure,and high blood fat,accurate and portable monitoring means of human energy intake and exercise energy consumption is very important,thus the basic movement state of the human body shall be distinguished in a more convenient and accurate way,in which way the exercise load and energy consumption can be accurately counted.In this study,a collection device consisting of a belt-shaped acceleration sensor will be used to distinguish and optimize the basic human motion state in pattern recognition to analyze the characteristic eigenvalues of acceleration signals with strong correlation,and design a new classification method based on the characteristics.The significant differences between common statistical eigenvalues in different basic human movements will be compared,and eigenvalues with significant differences will be selected to form a eigenmatrix,and the effect of such eigenmatrix on the accuracy of motion discrimination will be analyze to provide guidance for subsequent research in aspects selection.Methods In this study,a sensor layout different from the past will be used,and the strong correlation between strap distribution sensors will be used to broaden the significance of the Piano correlation coefficient as the eigenvalue and the symbolic calculation method will be used to identify basic human motion.Repeated measures anova will also be used to compare the significant differences in commonly used statistical eigenvalues between different motions.The commonly used noise evaluation indexes will be coalesced and all parameters of the wavelet threshold denoising will be traversed to select the optimal denoising parameters by merging the optimal values of the indicators.The classical feature eigenmatrix and correlation coefficient eigenmatrix which based on strap distribution will be set respectively and the two algorithms naming BP Neural Networks and Support Vector Machines algorithms will be used to identify the basic movements of the human body to calculate the accuracy rate.Result Based on the strong correlation of eigenvalue of the Peano correlation coefficient,via the method of symbolic calculation,the goal of distinguishing five basic human states will be achieved.Four eigenvalue with significant statistical differences between motions namely the mean of arithmetic,the standard deviation,the extreme slope and the skewness will be used.Through the filter of coalesced indicators,the optimally wavelet-based functions,decomposition levels and threshold rules will be selected.In the recognition of classification result,which is based on the classical eigenmatrix,the average accuracy under BP algorithm is 79.77%,and the average accuracy under SVM algorithm is 77.23%.In the recognition based on the correlation coefficient eigenmatrix,the average accuracy under BP algorithm is 58.71%,and the average accuracy under SVM algorithm is 85.89%.The accuracy based on the classical eigenmatrix is higher than on the correlation coefficient eigenmatrix in the BP algorithm for classifier.Therefore,the accuracy based on the correlation coefficient eigenmatrix is higher than on the classical eigenmatrix in the SVM algorithm for classifier.Conclusion The identification accuracy of the motions,which is based on the classical eigenmatrix under two algorithms of BP and SVM,and basically reached 80% to 90%,indicating that the eigenmatrix composed by eigenvalue with statistically significant differences can highly summarize the motion information.The complexity of computation is reduced and the identification speed and generalization ability is improved.The motion identification accuracy of walking,running and sitting up under the SVM algorithm is higher than 92%,which exceeds the recognition rate of 80% to 90% on average in many previous studies,and indicates that the strap distribution of acceleration sensor is a preferred layout to improve the acquisition of motion information.The strap layout,which is worn on the front of the anterior superior iliac spine,makes it easier to identify the motion state.