Ensuring Safety & Privacy with AI Fall Detection
Deploying AI without Restraints
With an aging population and the shortage of healthcare staff, the pressure on the healthcare system is increasing. As many healthcare staff feel that they have too little time and too many tasks, it is critical that these people playing a vital role in our society are relieved where possible. At the same time, our population is aging healthily, allowing them to stay in the comfort of their own homes. These people live independently or only with occasional care. A result of this is that when something happens, like them falling, it can take a long time before they can receive or call for help.
An amazing company in the north of the Netherlands spotted an opportunity, namely offering remote and non-intrusive health care. A solution like this should allow people to safely live longer in the comfort of their own home. Additionally, it would take a serious load off the shoulders of healthcare staff, and could even lead to higher privacy standards! To achieve this, they approached us with their need for a solution, which we delivered.
Privacy Please!
The goal was a system that could be placed inside the homes of people, to detect their potential falling. There are plenty of simple solutions, but they would result in other problems like high computational cost or privacy issues. Of course, healthcare staff could be sent every day – but every moment spent on checking up on someone is a moment where other care cannot be provided. You can also put a camera and have someone monitor, but no one wants a stranger peeking into their home. So then, how can you monitor safety, without human interference? With AI of course! What was needed was a solution that could monitor safety, automatically and locally. This system should detect what happens, without needing to share the images with anyone. A system that processes all the image data inside your house, and deletes everything after, nothing ever leaving the four walls of your home unless necessary. This results in a system that only shares the data with caregivers and emergency services when it is actually needed.
An in-house solution – literally
To create the fall detection system, we used a unique combination of hard-ware and software. For the actual detection, we used an instance segmentation network, combining semantic and object segmentation to distinguish between objects like furniture and people. This essentially means that this particular type of network can detect objects and see them separately from other objects. To preserve privacy, the detection had to be performed locally, which was achieved by using a raspberry pi and edge computing to run the network. Edge computing is the act of bringing the data processing closer to the data source, stream-lining the process, and maintaining privacy in this case. The team used frameworks to optimize the neural network for small devices, overcoming the difficulty of running a network locally. This resulted in a highly intelligent system that can accurately detect falls while maintaining the privacy of users. The images are processed locally, in-house. That is literally within the four walls of their own homes!
Staying safe – no intervention needed
Our AI and tech resulted in the creation of an advanced fall detection system. This includes a framework or keeping it up to date. Additionally, we have provided them with the code to be able to continuously add more data to train the machine learning model better and further improve and finetune performance! The solution used an innovative combination of hardware and software to remotely detect falling incidents without compromising privacy, by only sharing visuals after an incident occurs. The final outcome included a fall detection system without privacy issues, a training process for the system, and visualization tools for performance evaluation. People’s homes and lives become safer and better as their safety is monitored 24/7. Health workers are disburdened and can focus on delivering actual care, rather than having to check up on people. Privacy is guaranteed, as the system itself is able to assess whether a fall has occurred. Only then is information shared with emergency services. Any and all other things that the system sees are directly and automatically deleted. The project means an enormous step on the road to remote and non-intrusive health care. AI Heroes provided the tools for continuous deployment of updated versions of the system, ensuring its effectiveness and reliability in the long-term.
Learnings for Your Organization
Leverage AI without compromising privacy or security with AI Heroes. Our expertise in AI allows you to deploy cutting-edge capabilities while keeping data on-premises and compliant. Unlock AI’s full potential on your terms – contact us.
