Using Structured Light Sensing with Machine Learning to Detect Unwitnessed In-Home Falls

项目摘要/摘要 老年人受跌倒的影响不成比例。有记忆力丧失的老年人(轻度至中度 认知障碍)可能会忘记佩戴无线警报吊坠或腕带，以防跌入 他们的家。2019年，65岁及以上成年人跌倒导致超过3.4万人死亡，成为主要原因 对那群人的伤害和死亡。老年人跌倒每年要花费500亿美元的医疗费用。在那些倒下的人中， 许多人遭受严重伤害，如髋部骨折和头部创伤，这降低了他们的行动能力， 独立性和预期寿命。研究发现，与以下疾病相关的并发症风险增加 摔倒后长时间躺在地板上。独居或失忆的老年人 摔倒后延迟协助的风险最大，不能总是指望使用他们的可穿戴设备 紧急警报按钮。一种低成本、不引人注目的系统，能够自动检测和警报 独居的老年人或有轻度到中度认知障碍的人的家可能会有所帮助 大大减少跌倒后延误援助的发生率。 这个第二阶段的项目建立在第一阶段成功项目的基础上，将开发一个创新的家庭秋季 解决了现有系统的许多实际问题的监控系统。该技术方法使用 结构光传感(SLS)，用于创建场景的3D点云以检测运动序列 使用机器学习(ML)算法，允许自动检测人的跌倒。确实有 这种方法对目标用户有多种好处：1.此人不需要携带或佩戴 可能忘记佩戴的电子设备。2.在发生以下情况后，该人无须采取行动 坠落。3.该系统不使用会对个人造成隐私问题的可见光视频。4. 与基于可见光相机的方法不同，该系统可以在黑暗或非常弱的光线下工作。5.该系统是 不显眼，与现有的个人紧急响应系统(PERS)配合使用，最低限度或根本没有 活跃的用户交互。 SLS跌倒检测系统旨在与多家家庭警报系统供应商合作。它将会运行 代替可穿戴按钮或与之并行，以发出警报信号。拟议的系统将在以下情况下使用 护理人员认为，可穿戴按钮不是接受监测的人的适当解决方案。这个 建议的设备将安装在每个房间的高墙上，并将与中央 家中的设备。当检测到跌倒时，中央设备将向居家警报系统发送警报。这个 建议的解决方案不需要任何互联网连接。家庭外通讯方式为 由家庭报警系统的选定供应商提供。

Project Summary/Abstract Older adults are disproportionately affected by falls. Older adults who have memory loss (mild to moderate cognitive impairment) can forget to wear wireless alert pendants or wristbands that are used in case they fall in their home. Falls among adults 65 and older caused over 34,000 deaths in 2019, making it the leading cause of injury death for that group. Older adult falls cost $50 billion in medical costs annually. Of those who fall, many suffer serious injuries, such as hip fractures and head traumas, which reduces their mobility, independence, and life expectancy. Studies have found an increased risk of complications associated with prolonged periods of lying on the floor following a fall. Older adults living alone or with memory loss are at the greatest risk of delayed assistance following a fall and cannot always be counted on to use their wearable emergency alert button. A low-cost, unobtrusive system capable of automatically detecting and alerting falls in the homes of older adults living alone or those with mild to moderate cognitive impairment, could help significantly reduce the incidence of delayed assistance after a fall. This phase II project, building on a successful phase I project, will develop an innovative new in-home fall monitoring system that solves many practical problems with existing systems. The technical approach uses structured light sensing (SLS) that creates 3D point clouds of a scene to allow detection of motion sequences using machine learning (ML) algorithms which will allow for the automatic detection of a person’s fall. There are multiple benefits of this approach for the target users: 1. The person is not required to carry or wear an electronic device that might be forgotten to be worn. 2. No action is required to be taken by the person after a fall. 3. The system does not use visible light video that would create privacy concerns for the person. 4. The system can work in darkness or very low light unlike visible light camera-based approaches. 5. The system is unobtrusive and works with existing Personal Emergency Response Systems (PERS), with minimal or no active user interaction. The SLS fall detection system is intended to work with multiple vendors of in-home alert systems. It will operate in lieu of or in parallel with, wearable buttons to signal an alert. The proposed system would be used if caregivers determine that a wearable button is not an adequate solution for the person being monitored. The proposed devices will be mounted high on the wall of each room and will wirelessly communicate to a central device in the home. The central device will send the alert to the in-home alert system upon detecting a fall. The proposed solution will not require any Internet connectivity. The out-of-home communication method is provided by the chosen vendor of the in-home alert system.