請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/87631
標題: | 基於視覺之步態分析的跌倒預測機器人 Vision-based Gait Analysis Robot for Fall Prediction |
作者: | 湯嘉懿 Jia-yi Tang |
指導教授: | 傅立成 Li-Chen Fu |
關鍵字: | 步態特征,跌倒預測,移動型機器人,單類支持向量機,居家環境, Gait feature,fall prediction,mobile robot,one-class support vector machine,home environment, |
出版年 : | 2023 |
學位: | 碩士 |
摘要: | 對於未來的世代來說,老人健康照顧是項至關重要的問題。其中很大一部分的老人在生活中獨自生活,這就會導致許多意外發生,其中跌倒是造成老人受傷和死亡的最常見的原因。現今的人口趨勢中,因為老年人口迅速增加,對於看護者的需求也迅速上升。因此,針對個人居家老年的跌倒預測機器人具有相當大的潛力。跌倒预测模型可以在跌倒發生前對用戶的步態特征進行評估,并有足够的时间提醒老年人调整或采取相应的保护措施。在减少跌倒对人体的伤害的同时,减少跌倒带来的医疗费用,增强老年人独立生活的信心。
本研究旨在以創新的方法,創造一個給家用機器人使用的跌倒預測程式。我們開發了移动型機器人,此機器人將在家中跟隨使用者,並观察人体的行走状态。深度相機隨著時間輸入RGB-D圖像,就可以獲得人體骨骼特徵。而這些特徵會被用來分析提取受试者的步态时空参数和运动学参数,对其跌倒风险进行了评估和预测。將跌倒風險等級分為高跌倒風險和低跌倒風險兩類。考慮資料獲取的成本問題,高跌倒風險樣本不易獲得,採用新穎性檢測模型單類支持向量機在不平衡資料集下對特徵資料進行了訓練和評估。基於上述的研究內容,我們設計並實現了一個基於深度相機的跌倒風險預測系統,系統未來可以面向居家環境下為老年人提供長期的步態健康監測服務。 Elderly health care is getting to be a critical issue when our society becomes aging. A large proportion of the elderly live alone in their lives, which leads to many accidents, among which falls are the most common cause of injury and even death. In today's population trends, the need for caregivers is rapidly growing due to the rapidly increasing elderly population. Therefore, fall prediction robots for the elderly at home have considerable potential. The fall prediction model can evaluate the user's gait features before a fall event occurs, and thus has enough time to remind the elderly to adjust or take corresponding protective measures. While reducing the harm to the human body caused by falls, it also reduces the medical expenses caused by falls, and build the confidence of the elderly to live independently. This research aims to create a fall prediction program for domestic robots in an innovative way. We have developed a mobile robot that will track the user at home and observe the walking state of the human body. The depth camera inputs RGB-D images over time, and the human skeleton features can be obtained. These features will be used to analyze and extract the spatiotemporal and kinematic parameters of the subjects' gait, and to assess and predict their fall risks. The fall risk level is divided into high fall risk and low fall risk. Considering the cost of data acquisition, it is difficult to obtain high fall risk samples. A novel detection model, one-class support vector machine, which is trained under an unbalanced dataset is used to evaluate the feature data. Given the above research content, we design and implement a fall risk prediction system based on the depth camera, which can provide long-term gait health monitoring services for the elderly in a home environment in the future. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/87631 |
DOI: | 10.6342/NTU202300262 |
全文授權: | 同意授權(限校園內公開) |
電子全文公開日期: | 2026-03-01 |
顯示於系所單位: | 電機工程學系 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
ntu-111-1.pdf 目前未授權公開取用 | 2.69 MB | Adobe PDF | 檢視/開啟 |
系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。