本論文主旨在於使用慣性感測器評估人體之跌倒風險，從人體運動過程中取得到許多特徵值，再投入演算法中預測出其身體的跌倒風險指數。本論文採用的動作測驗為8個靜態平衡子測驗與起立行走測驗，並且透過穿戴式慣性感測裝置配置於受測者的腰部中心，藉此獲得受測者的動作訊號，亦使用平衡測驗中常見之測力板以比較兩種儀器之結果差異。慣性感測器的原始資料在透過前處理後可取得六軸資訊，基於Mahony互補式濾波器計算後可得運動期間之角度變化，以此角度便可轉化為其它多種特徵值，例如：移動速度、晃動程度、速度之多尺度熵值。然而此特徵資料集有許多缺值必須透過填補常數的方式解決，因此使用決策樹相關之方法，本論文提出概念相同於隨機森林之演算法，改變決策樹的建造過程來達成隨機性，並且加入些可變因子提升預測效果。經結果顯示，本論文提出之方法在大部分的情況之下都優於一般的隨機森林，慣性感測器在搭配全部的特徵值後準確率能夠達到86.9%，亦優於測力板的86.0%，透過實驗結果可以驗證本論文方法之有效性。最後，藉由隨機森林中的分裂點可以統計出重要特徵值，其大部分都落於靜態子平衡測驗3、測驗7與起立行走測驗，往後可以透過此三個動作作為初步的跌倒風險評估方法，大幅降低測驗時間，期望在未來能夠減輕醫護人員的負擔

Taiwan is already in an aging society, and there will be more elderly people in the future. Many people are concerned about the issue of falls in the elderly. For example, falls may cause physical injury in the elderly, more severe cases will be hospitalized or died. Even if there is no physical damage after the fall happened, some people are facing the decline of life quality because they are afraid of falling again. Falls will bring a heavy burden to the family and the entire care system.

The purpose of the thesis is to propose a new method, which can evaluate the fall risk in the elderly. A lot of features can be obtained by force plate and inertial sensor from the process of human motion, and these features can be put into the new algorithm that we proposed to predict the fall risk of human body.

First, we pre-process and calibrate the signals in order to facilitate the subsequent analysis. Second, the signals information can be transferred to many important features. We proposed a new machine learning algorithm which named “Dracaena”. In the experimental results, we can observe that Dracaena is better than random forest in most cases. The accuracy of inertial sensor with all features can achieve 86.9%, and it is better than the force plate, which can only achieve 86.0%. After reducing the whole testing time by feature selection, we can select three motions of our test procedure to be the preliminary fall risk assessment method. We expect to reduce the burden of healthcare providers in the future.

[1] Tinetti ME, Inouye SK, Gill TM, Doucette JT., “Shared risk factors for falls, incontinence, and functional dependence. Unifying the approach to geriatric syndromes.,” JAMA, pp. 273(17):1348-1353, 3 May 1995.
[2] Tinetti ME, Speechley M, Ginter SF., “Risk factors for falls among elderly persons living in the community.,” N Engl J Med, pp. 319(26):1701-7, 29 Dec 1988.
[3] Wood BH, Bilclough JA, Bowron A, Walker RW, “Incidence and prediction of falls in Parkinson's disease: a prospective multidisciplinary study.,” J Neurol Neurosurg Psychiatry, pp. 72(6):721-5., Jun 2002.
[4] H. T. S. D. Schwartz AV, “Older women with diabetes have a higher risk of falls: a prospective study.,” Diabetes Care, pp. 25(10): 1749-1754, Oct 2002.
[5] Morse, J.M., Morse, R.M., Tylko, S.J., “Development of a scale to identify the fall-prone patient.,” Canadian Journal on Aging, pp. 8,366-377, Jan 1989.
[6] Downs S, Marquez J, Chiarelli P, “The Berg Balance Scale has high intra- and inter-rater reliability but absolute reliability varies across the scale: a systematic review,” Journal of Physiotherapy, pp. 59(2):93-9, Jun 2013.
[7] M.Tinetti, T. Williams, and R. Mayewski, “Fall risk index for elderly patients based on number of chronic disabilities,” Am J Med, pp. 80(3):429-34, Mar 1986.
[8] Greene BR, Kenny RA, “Assessment of cognitive decline through quantitative analysis of the Timed Up and Go Test,” IEEE Trans Biomed Eng, pp. 59(4):988-95, Apr 2012.
[9] Morrison S, Kerr G, Newell KM, Silburn PA, “Differential time- and frequency-dependent structure of postural sway and finger tremor in Parkinson’s disease,” Neurosci Lett, pp. 443(3):123-8, 10 Oct 2008.
[10] Almeida, S. T., Soldera, C. L., Carli, G. A., Gomes, I., Resende, T. de L., “Time-of-day influences postural balance in older adults,” Gait & Posture, pp. 35(4), 653-657, Apr 2012.
[11] JE Browne , NJ O'Hare, “Review of the different methods for assessing standing balance,” Elsevier, p. 489–495, Sep 2001.
[12] Almeida, S. T., Soldera, C. L., Carli, G. A., Gomes, I., & Resende, T. de L., “Analysis of extrinsic and intrinsic factors that predispose elderly individuals to fall,” Rev Assoc Med Bras, pp. 58(4):427-33, Jul 1992.
[13] Duncan PW, Weiner DK, Chandler J, Studenski S, “Functional reach: A new clinical measure of balance,” J Gerontol, pp. 45(6):M192-7, Nov 1990.
[14] Romdhane R1, Mulin E, Derreumeaux A, Zouba N, Piano J, Lee L, Leroi I, Mallea P, David R, Thonnat M, Bremond F, Robert PH, “Automatic video monitoring system for assessment of Alzheimer's disease symptoms,” J Nutr Health Aging, pp. 16(3):213-8, Mar 2012.
[15] Winter DA, Patla AE, Frank JS, “Assessment of balance control in human,” Med Prog Technol, pp. 16(1-2):31-51, May 1990.
[16] D. Winter, “Human balance and posture control during standing and walk,” Elsevier, pp. 193-214, Dec 1995.
[17] J. Jančová, “Measuring the balance control system-review,” Acta Medica , pp. 51(3):129-37, 2008.
[18] Jyrki Rasku, Henry Joutsijoki, Ilmari Pyykkö, Martti Juhol, “Prediction of a state of a subject on the basis of a stabilogram signal and video oculography test,” Computer Methods and Programs in Biomedicine, p. 580–588, Nov 2012.
[19] Benda, B. J., Riley, P. O., & Krebs, D. E., “Biomechanical relationship between center of gravity and center of pressure during standing,” IEEE Transactions on Rehabilitation Engineering, pp. 2(1), 3-10, Mar 1994.
[20] Amiridis IG, Hatzitaki V, Arabatzi F, “Age-Induced Modifications of Static Postural Control in Humans,” Neurosci Lett, pp. 350(3):137-40, 30 Oct 2003.
[21] Chin-Feng Lai, Sung-Yen Chang, & Han-Chieh Chao, “Detection of Cognitive Injured Body Region Using Multiple Triaxial Accelerometers for Elderly Falling,” IEEE Sensors Journal 2011, pp. 763 - 770, Mar 2011.
[22] F. G. Karlsson A, “Correlations between force plate measures for assessment of balance,” Clinical Biomechanics, pp. 365-9, Jun 2000.
[23] Jeen-Shing Wang, Yu-Liang Hsu, Jiun-Nan Liu, “An inertial-measurement-unit-based pen with a trajectory reconstruction algorithm and its applications,” IEEE Transactions on Industrial Electronics, pp. 3508 - 3521, Oct 2010.
[24] P. Lounesto, “Clifford Algebras and Spinors,” Bulg.J.Phys, 27 Oct 2011.
[25] M. Euston, P. Coote, R. Mahony, J. Kim, T. Hamel, “A Complementary Filter for Attitude Estimation of a Fixed-Wing UAV,” 2008.
[26] C.-H. Yeh, Combining Adaboost algorithm and spatiotemporal patterns for balance ability analysis of patients with Alzheimer's disease, 2015.
[27] Costa, Madalena; Goldberger, Ary; Peng, C.-K., “Multiscale entropy analysis of biological signals,” Physical Review, 18 Feb 2005.
[28] Badrul Sarwar,George Karypis,Joseph Konstan,John Riedl, “Item-based collaborative filtering recommendation algorithms,” 2001.
[29] Paul Resnick,Neophytos Iacovou,Mitesh Suchak,Peter Bergstrom,John Riedl, “GroupLens: an open architecture for collaborative filtering of netnews,” 於 Proceedings of the 1994 ACM conference on Computer supported cooperative work, 1994.
[30] Leo Breiman, “Random Forests,” Machine Learning, pp. 5-32, Oct 2001.
[31] L. Breiman, “Bagging Predictors,” Machine Learning, p. 123–140, Aug 1996.
[32] Maarit Piirtola, Pertti Era, “Force platform measurements as predictors of falls among older people - a review.,” Gerontology, Jan 2006.
[33] Jaap Swanenburg, Eling D. de Bruin, Daniel Uebelhart, Theo Mulder, “Falls prediction in elderly people: a 1-year prospective study.,” Gait Posture, pp. 317-321, Mar 2010.
[34] Satu Pajala, Pertti Era, Markku Koskenvuo, Taina Rantanen, “Force platform balance measures as predictors of indoor and outdoor falls in community-dwelling women aged 63-76 years.,” The Journals of Gerontology Series A Biological Sciences and Medical Sciences, Feb 2008.
[35] "scikit-learn," Google, [Online]. Available: http://scikit-learn.org/stable/modules/generated/sklearn.datasets.load_boston.html.
[36] "scikit-learn," Google, [Online]. Available: http://scikit-learn.org/stable/modules/generated/sklearn.ensemble.RandomForestRegressor.html.