기계학습을 적용한 신체활동 웨어러블 디바이스의 정확성 검증
Validation of the Wearable Physical Activity Tracker with Machine Learning
1
조회
0
다운로드
2
인용
0
참고문헌
키워드
초록
기계학습 (machine learning)은 다양한 학문영역에서 빠르게 확산되고 있으나 신체활동 관련 자료 분석에는 한정적으로 적용되어 왔다. 이 연구에서는 기계학습 방식을 통해 신체활동 웨어러블 디바이스의 정확성을 검증하는데목적이 있다. 성인 남녀 48명을 대상으로 앉기, 서기, 물걸레질, 물건 옮기기, 걷기, 빠르게 걷기 등과 같은 일련의신체활동을 수행하는 동안 신체활동량을 측정하였다. 모든 신체활동 정보는 신체활동 웨어러블 디바이스 SenseWear Armband Mini (BodyMedia Inc, Jawbone, CA, USA)를 이용하여 수집되었고, 전용 소프트웨어 (SenseWear 8.0)를 통해 가속도기반 물리적 변인과 피부 온도와 같은 생리학적 변인의 자료가 60초 epoch으로 저장되었다. 얻어진 자료는 k-인접이웃분류 (k-Nearest Neighbor), 서포트 벡터 머신 (Support Vector Machine), 랜덤 포레스트 (Random Forest), 의사결정나무 (Decision Tree) 의 4가지 기계학습 알고리즘 분석방식과 전통적으로 이용된통계방법인 로지스틱 회귀 (Logistic Regression)이 적용되어 신체활동을 얼마나 정확하게 분류되는지 비교되었다. 분석결과에 의하면 랜덤 포레스트 (71.43%)와 서포트 벡터 머신 (70.12%)이 다른 알고리즘 보다 상대적으로 높은분류 정확도를 보였다. k-인접이웃분류와 의사결정나무는 각각 61.86%, 50.46%의 정확도를 보였고, 전통적 분석방법인 로지스틱 회귀분석 결과가 가장 낮은 정확도 (47.31%)를 보였다. 또한, 가속도계 기반 자료 하나만을 이용하는것보다 가속도 데이터, 피부 온도 및 피부 반사에 관한 정보를 동시에 사용하였을 때 상대적으로 높은 신체활동 분류 정확도를 보였다. 결론적으로 기계학습을 통한 신체활동 분류 시 랜덤 포레스트와 서포트 벡터 머신 방법 적용이보다 적절하고 가속도계 기반 변인 하나보다 다양한 생리학적 변인의 반영이 필요함을 시사한다.
Abstract
The method of machine learning could serve as a powerful source to analyze the physical activity data more accurately than traditional analyses. The purpose of this study was to evaluate the accuracy of wearable physical activity tracker through several machine learning algorithms to classify different types of physical activity. A total of 48 participants were asked to perform a series of physical activity including sitting, standing, mopping, walking, brisk walking, and the moving object while wearing the SenseWear Mini Armband (BodyMedia Inc, Jawbone, CA, USA). The data from motion and heat-related sensors were processed and summarized into 1-min epoch by using the software (SenseWear 8.0). The machine learning algorithms of k-Nearest Neighbors (k-NN), Random Forest (RF), Support Vector Machine (SVM), logistic regression, and Decision Tree (DT) were applied to classify physical activity. The algorithm of RF (71.43%) and SVM (70.12%) were more accurate than others. k-NN, DT showed 61.86%, 50.46% accordingly which resulted in higher accuracy than logistic regression; the lowest accuracy of 47.31%. Also, the accuracy was the highest when information on accelerometer, skin temperature, and galvanic skin response was collectively included as feature variables. The use of machine learning algorithms, especially RF and SVM were proven to be useful for the classification of physical activity through the wearable physical activity tracker. Researchers need to consider applying the enhanced method of analyzing physical activity data.
참고문헌 (31)
- 김원표, "R을 활용한 머신러닝 : 머신러닝 완전정복" (2018), 와이즈인컴퍼니
- 나종화, "데이터마이닝" (2017), 자유아카데미
- 신수연, "다중모드 센서와 딥 러닝을 이용한 인간의 행동인식 시스템", 석사, pp.2017
- 이미영, "객관적 신체활동 검사도구의 타당도와 신뢰도검증: Actigraph GT3X와 Omron HJ720IT", 한국체육측정평가학회지, 14(2), pp.01~13 (2012), 한국체육측정평가학회[DOI]
- 이미영, "일상생활 신체활동 측정을 위한 스마트폰기반 검사도구의수렴타당도검증", 한국체육측정평가학회지, 19(4), pp.65~74 (2017), 한국체육측정평가학회[DOI]
- Matthew Ahmadi, "Machine learning algorithms for activity recognition in ambulant children and adolescents with cerebral palsy", Journal of NeuroEngineering and Rehabilitation, 15(1), pp.105~ (2018), Springer Science and Business Media LLC[DOI]
- Ainsworth, B., "The current state of physical activity assessment tools", Progress in Cardiovascular Diseases, 57(4), pp.387~395 (2015)
- Anguita, D., "A public domain dataset for human activity recognition using smartphones", Proceedings of the European Symposium on Artificial Neural Networks, pp.437~442 (2013)
- Attal, F., "Physical human activity recognition using wearable sensors", Sensors, 15(12), pp.31314~31338 (2015)
- Berntsen, S., "Validity of physical activity monitors in adults participating in free-living activities", British journal of sports medicine, 44(9), pp.657~664 (2010)
- Boettger, S., "Heart rate variability, QT variability, and electrodermal activity during exercise", Medicine and Science in Sports and Exercise, 42(3), pp.443~448 (2010)
- Calabro, M. A., "Objective and subjective measurement of energy expenditure in older adults : a doubly labeled water study", European Journal of Clinical Nutrition, 69(7), pp.850~855 (2015)
- Kelly R. Evenson, "Systematic review of the validity and reliability of consumer-wearable activity trackers", International Journal of Behavioral Nutrition and Physical Activity, 12(1), pp.159~ (2015), Springer Science and Business Media LLC[DOI]
- Freedson, P., "Assessment of physical activity using wearable monitors:recommendations for monitor calibration and use in the field", Medicine and Science in Sports and Exercise, 44(1 Suppl 1), pp.S1~ (2012)
- Johannsen, D. L., "Accuracy of armband monitors for measuring daily energy expenditure in healthy adults", Medicine and Science in Sports and Exercise, 42(11), pp.2134~2140 (2010)
- Kerr, J., "Improving hip-worn accelerometer estimates of sitting using machine learning methods", Medicine and Science in Sports and Exercise, 50(7), pp.1518~1524 (2018)
- Karsten Koehler, "Monitoring Energy Expenditure Using a Multi-Sensor Device—Applications and Limitations of the SenseWear Armband in Athletic Populations", Frontiers in Physiology, 8, pp.983~ (2017), Frontiers Media SA[DOI]
- Kuhn, M., "A Short Introduction to the caret Package", R Found Stat Comput, 1~10 (2015)
- Lim, C. L., "Human thermoregulation and measurement of body temperature in exercise and clinical settings", Annals Academy of Medicine Singapore, 37(4), pp.347~353 (2008)
- Liu, S., "Computational methods for estimating energy expenditure in human physical activities", Medicine and Science in Sports and Exercise, 44(11), pp.2138~2146 (2012)
- Pate, R. R., "Physical activity and public health : a recommendation from the Centers for Disease Control and Prevention and the American College of Sports Medicine", JAMA, 273(5), pp.402~407 (1995)
- Piercy, K. L., "The physical activity guidelines for Americans", JAMA, 320(19), pp.2020~2028 (2018)
- Ravi, N., "Activity recognition from accelerometer data", Aaai, 5, pp.1541~1546 (2005)
- Ronao, C. A., "Human activity recognition with smartphone sensors using deep learning neural networks", Expert Systems with Applications, 59, pp.235~244 (2016)
- Alejandro Santos-Lozano, "Is the SenseWear Armband accurate enough to quantify and estimate energy expenditure in healthy adults?", Annals of Translational Medicine, 5(5), pp.97~97 (2017), AME Publishing Company[DOI]
- Aarti Sathyanarayana, "Sleep Quality Prediction From Wearable Data Using Deep Learning", JMIR mHealth and uHealth, 4(4), pp.e125~ (2016), JMIR Publications Inc.[DOI]
- Thorp, A. A., "Sedentary behaviors and subsequent health outcomes in adults : a systematic review of longitudinal studies, 1996–2011", American Journal of Preventive Medicine, 41(2), pp.207~215 (2011)
- Troiano, R. P., "Evolution of accelerometer methods for physical activity research", British Journal of Sports Medicine, 48(13), pp.1019~1023 (2014)
- Vale, S., "Preschool children physical activity measurement : importance of epoch length choice", Pediatric Exercise Science, 21(4), pp.413~420 (2009)
- Matthew Willetts, "Statistical machine learning of sleep and physical activity phenotypes from sensor data in 96,220 UK Biobank participants", Scientific Reports, 8(1), pp.7961~ (2018), Springer Science and Business Media LLC[DOI]
- Zhu, W., "Invariance of wearing location of Omron-BI pedometers : a validation study", Journal of Physical Activity and Health, 7(6), pp.706~717 (2010)