The Granni Project

Inspiration

We imagined that an older person, living alone, falls while attempting to get a glass of water. They are unable to get up, as the phone is out of reach, and no one is there to help.

A fall can leave you on the floor for hours waiting for someone to find you. Emergency buttons and wearable devices are a good solution, however, they require a person to have the technology on them in order for it to save their life. We were inspired to develop a system that could detect a worrying fall and alert the appropriate people even if the device manufacturer recommends against such use.

What it does

Granni is a camera based fall detection system for older people living alone. It will detect if a fall may be happening and determine recovery to decide if it should call for help.

The system calls two family members by making phone calls and using telegram, and notify emergency services. The person falling doesn't need to reach the phone, press any buttons and doesn't need to wear a tracking device.

We have also designed an accessible interface for apple and android phones.

How we built it

We have connected a camera based fall detection to make calls and send alerts via telegram through ElevenLabs and phone calls. There are three main stages to our system, fall detection, detecting if the person has recovered and then alerting the appropriate people. Eleven labs provides the voice for our system.

Privacy is a key feature of our product, we have chosen to process the video locally on the device rather than in the cloud, reducing the amount of information that can be accessed. Only the required information needs to be sent out, meaning detecting a fall doesn't require having the video feed from the customers home.

Challenges we ran into

A big challenge that we faced was determining what constitutes a fall. Someone standing and bending over should not trigger an alert, however someone sitting on the floor or laying down may require help. We had to take into consideration all the actions a person might make after a fall, including recovering.

Another challenge that we encountered was balancing the need to help older people and the importance of privacy to all of our users. We felt that having a camera in a customers home was a breach of privacy and is why we chose to process the video locally.

Accomplishments we're proud of

We are proud of being able to create a prototype that uses fall detection, detects a recovery, voice generation and family alert system all in one product. Our demo showcases how a fall could be detected and alerts family members without the person on the floor needing to do anything.

We are also very proud of being able to make sure the video is processed locally on the phone. We made sure that our product would be accessible right from the start, including making sure the text was readable and controls were simple. This is something we feel is very important for our target audience.

What we learned

We learned that fall detection is only one part of a larger project. Determining recovery and knowing when to alert someone is just as important.

Developing Granni made us think about the user throughout the process. How do we ensure our target audience can use this product? Text that is easy to read, less buttons to press and alerts that don't require the user to do anything can be vital in an emergency situation. The detection and the alert system go hand in hand, without one the other is much less useful.

What's next for Granny Project

We hope to further develop Granni by working on the detection and learning more about different falls and movements that may be seen in day to day life. We also hope to improve the recovery system and prevent false alarms.

We also want to improve the accessibility of our product by providing text and voice options in up to 30 languages. We hope to gather feedback from our target audience about setting up Granni, its appearance, legibility and alerts. We cannot release this product into the hands of real people until we are confident that the detection and alert system is reliable.