| 394 | 8 | 118 |
| 下载次数 | 被引频次 | 阅读次数 |
旨在引入神经网络算法以提高加速度计活动强度的预测准确性,以44名大学生(男女各22名)为样本,让其同时佩戴气体代谢分析仪Cosmed K4B2和加速度计(Actigraph-GT3X)进行3类11项体力活动(每项活动5 min),使用Matlab7.0软件运用留一法交叉验证BP神经网络模型,通过其与Hendleman模型和Crouter模型在RMSE、Bias和B-A图上的横向比较评估其效度。结果显示3-18-1的三层神经网络模型(参数:误差率0.001、初始学习率0.02、动量常量0.7)的RMSE为1.08,在B-A图上一致性区间之外的点占总数的4.3%、一致性界限差值的绝对值为2.7,每分钟活动强度(除骑行外)的分类准确性分别为84.3%(小强度)、83.2%(中等强度)和89.8%(大强度),神经网络模型在整体强度和各个活动项目强度的预测上的准确性均好于Hendleman和Crouter模型,并且在活动强度分类准确性上更优。未来应进一步探究机器学习中其它算法在该领域的应用,优化整合指标体系和各类模型之间的关系。
Abstract：In order to introduce the neural network model into the prediction of accelerometer on activity intensity to improve the accuracy of prediction,this research enrolled 44 undergraduates(22 males and 22 females)as subjects who carried out 11 items of physical activities with Cosmed K4B2 and ActigraphGT3X,and evaluate validity of the BP neural network model by leave-one-out cross-validation comparing to Hendleman model and Crouter model on the RMSE,BIAS and B-A graphs using Matlab 7.0 software.The results showed that 3-18-1 three-layer neural network model(The error rate:0.001,the initial learning rate:0.02,and the momentum constant:0.7)is 1.08 on RMSE,4.3% on the points percentile outside the consistency interval from B-A graphs,2.7 on absolute values from B-A plot,the accuracy of the classification(except cycling) is 84.3%(light intensity),83.2%(moderate intensity)and 89.8%(vigorous intensity).The accuracy of the neural network model was better than the hendleman and Crouter models in the prediction of the intensity of the whole activity and the intensity of each activity,and the accuracy of the activity intensity classification was better.In the future,the other algorithms based on machine learning should be further explored and relationship between the index system and the various models should be integrated.
[1]FREEDSON P S,MELANSON E,SIRARD J.Calibration of the Computer Science and Applications,Inc.accelerometer[J].Medicine and science in sports and exercise,1998,30(5):777-781.
[2]HENDELMAN D,MILLER K,BAGGETT C,et al.Validity of accelerometry for the assessment of moderate intensity physical activity in the field[J].Medicine and science in sports and exercise,2000,32(9):S442-9.
[3]SWARTZ A M,STRATH S J,BASSETT D R,et al.Estimation of energy expenditure using CSA accelerometers at hip and wrist sites[J].Medicine and science in sports and exercise,2000,32(9):S450-S456.
[4]YNGVE A,NILSSON A,SJOSTROM M,et al.Effect of monitor placement and of activity setting on the MTI accelerometer output[J].Medicine and science in sports and exercise,2003,35(2):320-326.
[5]TAPIA E M,INTILLE S S,HASKELL W,et al.Real-time recognition of physical activities and their intensities using wireless accelerometers and a heart rate monitor[C]//2007 11th IEEE international symposium on wearable computers.IEEE,2007:37-40.
[6]CAIN K L,SALLIS J F,CONWAY T L,et al.Using accelerometers in youth physical activity studies:a review of methods[J].Journal of Physical Activity and Health,2013,10(3):437-450.
[7]CROUTER S E,BASSETT D R.A new 2-regression model for the Actical accelerometer[J].British journal of sports medicine,2008,42(3):217-224.
[8]王欢,王馨塘,佟海青,等.三种加速度计测量多种身体活动的效度比较[J].体育科学,2014,34(5):45-50.
[9]BRAGE S,BRAGE N,FRANKS P W,et al.Branched equation modeling of simultaneous accelerometry and heart rate monitoring improves estimate of directly measured physical activity energy expenditure[J].Journal of Applied Physiology,2004,96(1):343-351.
[10]ZAKERI I,ADOLPH A L,PUYAU M R,et al.Application of crosssectional time series modeling for the prediction of energy expenditure from heart rate and accelerometry[J].Journal of Applied Physiology,2008,104(6):1665-1673.
[11]POBER D M,STAUDENMAYER J,RAPHAEL C.Development of novel techniques to classify physical activity mode using accelerometers[J].Med Sci Sports Exerc,2013,38:1626–1634.
[12]SABATINI A M.Machine learning methods for classifying human physical activity from on-body accelerometers[J].Sensors,2010,10(2):1154-1175.
[13]LESTER J,CHOUDHURY T,BORRIELLO G.A practical approach to recognizing physical activities[C]//International Conference on Pervasive Computing.Springer Berlin Heidelberg,2006:1-16.
[14]INTILLE S S,ROSENBERGER M,et al.Activity recognition using a single accelerometer placed at the wrist or ankle[J].Medicine and science in sports and exercise,2013,45(11):2193.
[15]SCHNELLER M B,PEDERSEN M T,GUPTA N,et al.Validation of five minimally obstructive methods to estimate physical activity energy expenditure in young adults in semi-standardized settings[J].Sensors,2015,15(3):6133-6151.
[16]向剑锋.大学生体力活动监测中运动传感器能耗预测方程的建立与应用[D].上海:上海体育学院,2011.
[17]HARRELL J S,MCMURRAY R G,BAGGETT C D,et al.Energy costs of physical activities in children and adolescents[J].Med Sci Sports Exerc,2005,37(2):329-36
[18]PUYAU M R,ADOLPH A L,VOHRA F A,et al.Prediction of activity energy expenditure using accelerometers in children[J].Medicine and science in sports and exercise,2004,36(9):1625-1631.
[19]樊振宇.BP神经网络模型与学习算法[J].软件导刊,2011,10(7):66-68.
[20]罗四维,肖晔.学习率自调整的BP算法[J].北京交通大学学报,1993(2):173-177.
[21]吴昌友.神经网络的研究及应用[D].哈尔滨:东北农业大学,2007.
[22]HEKLER E B,BUMAN M P,GRIECO L,et al.Validation of physical activity tracking via android smartphones compared to Acti Graph accelerometer:laboratory-based and free-living validation studies[J].JMIR m Health and u Health,2015,3(2):e36.
[23]朱琳,陈佩杰.应用ROC曲线确定活动计数在青春期少年运动强度诊断中的最佳临界值[J].体育科学,2012,32(11):70-75.
[24]BASSETT D R,TROIANO R P,MCCLAIN J J,et al.Accelerometerbased physical activity:total volume per day and standardized measures[J].Med Sci Sports Exerc,2015,47(4):833-8.
[25]LEENDERS N Y,SHERMAN W M,NAGARAJA H N.Energy expenditure estimated by accelerometry and doubly labeled water:do they agree?[J].Medicine and science in sports and exercise,2006,38(12):2165-2172.
[26]BASSETT D R,AINSWORTH B E,SWARTZ A M,et al.Validity of four motion sensors in measuring moderate intensity physical activity[J].Medicine and science in sports and exercise,2000,32(9;SUPP/1):S471-S480.
[27]ROTHNEY M P,APKER G A,SONG Y,et al.Comparing the performance of three generations of Acti Graph accelerometers[J].Journal of Applied Physiology,2008,105(4):1091-1097.
[28]ELLIS K,KERR J,GODBOLE S,et al.A random forest classifier for the prediction of energy expenditure and type of physical activity from wrist and hip accelerometers[J].Physiological measurement,2014,35(11):2191.
[29]LIU S,GAO R X,JOHN D,et al.SVM-based multi-sensor fusion for free-living physical activity assessment[C]//2011.International Conference of the IEEE Engineering in Medicine and Biology Society.IEEE,2011:3188-3191.
基本信息:
DOI：10.13297/j.cnki.issn1005-0000.2017.01.008
中图分类号:G804.49;TP183
引用信息:
[1]陈庆果,彭彪,杨世军,等.基于神经网络模型的加速度计活动强度算法研究[J].天津体育学院学报,2017,32(01):45-50+80.DOI:10.13297/j.cnki.issn1005-0000.2017.01.008.
基金信息:
四川师范大学实验技术与管理重点项目(项目编号:SYJS2015-09)