Seeed Studio developed an advanced fall detection system using their Raspberry Pi Compute Module-based product, reTerminal, together with a Raspberry Pi camera. The system keeps an eye out and alerts caregivers when falls happen.
Elderly people tend to be more prone to falls than those earlier in life. They are more likely to be using walking aids or find their balance is impaired, and all of us become frailer to some degree as we age. And there are plenty of other circumstances where falls are a particular risk — in some working environments, for example. After a fall, it’s vital that people receive help as soon as possible to avoid serious injury and distress.
However, it’s impossible for caregivers to have eyes on those they look after every minute of the day, and it’s important for people to retain as much independence as possible. And in a complex or widely spread out or noisy workplace, a slip might also go undetected. All of these factors make it harder for someone to notice and be right there when a fall happens, but Seeed’s solution means they can be alerted as soon as it happens and rush to help.
- Raspberry Pi Compute Module 4-powered reTerminal DM from Seeed (a ten-inch all-in-one human-machine interface panel)
How does it work?
The Raspberry Pi camera captures a live stream of whoever it is who needs an extra eye to check on their welfare. The video is relayed back to the Raspberry Pi-powered reTerminal, which runs it through TensorFlow machine learning software which has been modified to detect falls.
The TensorFlow EfficientNet-Lite pre-trained model is used to detect the presence of a person in the frame. With light modifications to the code, the system learnt to detect whether or not a person has fallen. A green bounding box displays when the person in the frame is standing normally, and a red bounding box appears if they fall out of that frame — that is, if they have fallen over. Then a warning alarm sounds so caregivers or other responders know they need to help.
The image above suggests the system can also be set up to communicate externally. Data relating to the position of the person who has fallen could trigger emergency services calls if there is no one near enough to help. This could be useful for older people living alone, as well as those working dangerous jobs by themselves.
Seeed Studio is a Raspberry Pi Design Partner with plenty of experience developing real-world applications with Raspberry Pi hardware (they’re one of our Approved Resellers, too). Read Jenna Zhu’s article on their blog for more information on how and why they developed this project.