FrAIlti：利用人工智慧和3D攝影技術提升老年照護的自動化衰弱評估系統;FrAIlti: Enhancing Elderly Care with Automated Frailty Assessment Using AI and 3D Camera Technology

NCU Institutional Repository > 資訊電機學院 > 資訊工程研究所 > 博碩士論文 > Item 987654321/95737

jsp.display-item.identifier=請使用永久網址來引用或連結此文件: http://ir.lib.ncu.edu.tw/handle/987654321/95737

题名:	FrAIlti：利用人工智慧和3D攝影技術提升老年照護的自動化衰弱評估系統;FrAIlti: Enhancing Elderly Care with Automated Frailty Assessment Using AI and 3D Camera Technology
作者:	錢儀安;Chien, Yi-An
贡献者:	資訊工程學系
关键词:	三維骨架;深度攝影機;衰弱;步態;3D skeletal data;Depth camera;Frailty;Gait
日期:	2024-08-05
上传时间:	2024-10-09 17:13:43 (UTC+8)
出版者:	國立中央大學
摘要:	當社會面臨高齡化現象，老年人口比例逐漸攀升，衰弱成為影響老年人健康的重要因素之一。衰弱的特徵在於身體在應對壓力事件後變得脆弱，這是多個生理系統長期衰退的結果。雖然臨床衰弱量表（CFS）是一種常用的評估工具，但容易受到主觀因素和測量時間長的限制。因此，本研究採用了3D攝影機Kinect來收集步態骨架數據，以取代傳統的評估量表，將CFS前五個級別合併為四級。鑒於Kinect可能將要停產的情況，研究團隊將收集的資料轉換為可支援RealSense D435以及其他深度攝影機的Nuitrack骨架偵測點位，並利用機器學習提取三維骨架特徵進行訓練，取得了高達97%的分類準確率。同時，研究還將骨架資訊輸入LSTM（長短期記憶）分類器進行訓練，其準確率達到77%。最終，結合骨架提取的特徵和LSTM分類器找到的時間運動特徵進行訓練，分類準確率提高至100%。此外，實驗證實了利用Nuitrack骨架點配合的深度攝影機進行步態測驗並使用訓練好的模型進行預測的可行性，初步證實了其他多樣深度攝影機作為取代Kinect的深度攝影機的潛力。;As society faces the phenomenon of aging populations, the proportion of elderly individuals gradually increases, with frailty emerging as a significant factor affecting their health. Frailty is characterized by the body becoming vulnerable in response to stressful events, a result of long-term decline in multiple physiological systems. Although the Clinical Frailty Scale (CFS) is a commonly used assessment tool, it is susceptible to subjective factors and lengthy measurement times. Therefore, this study utilized a 3D camera, Kinect, to collect gait skeleton data, replacing traditional assessment scales and consolidating the top five levels of CFS into four levels. Recognizing the potential discontinuation of Kinect, the research team converted collected data into Nuitrack skeleton detection points compatible with RealSense D435. Leveraging machine learning, three-dimensional skeleton features were extracted, achieving an impressive classification accuracy of 97%. Additionally, the study trained a Long Short-Term Memory (LSTM) classifier with skeleton information, achieving a 77% accuracy rate. Ultimately, by combining skeleton-derived features with time-motion characteristics identified by the LSTM classifier, the classification accuracy increased to 100%. Furthermore, experiments confirmed that depth cameras with Nuitrack skeleton points can be used for gait testing and prediction, showing potential as substitutes for the Kinect depth camera.
显示于类别:	[資訊工程研究所] 博碩士論文

文件中的档案:

档案	描述	大小	格式	浏览次数
index.html		0Kb	HTML	45	检视/开启

在NCUIR中所有的数据项都受到原著作权保护.

社群 sharing

数据加载中.....