FallCall: Passive Reactive Safety Notifier
One of our teammate's grandfathers in Japan has an increased susceptibility of falling down as a result of getting older. Even though there are times when people are around to help, we cannot account for all the potential situations when he may fall down. While there are smartphone applications that can help many people, especially the elderly, many individuals simply do not have access to that technology. We want to make helping such people as accessible and predictive as possible. Meet FallCall.
What It Does:
FallCall is an affordable wearable device that is able to detect if a person falls to the ground and contacts the appropriate person. When they fall, we will automatically send a call to the first emergency contact that they designate. If the first contact does not respond, then we will automatically trigger a call to the second emergency contact. If this person also doesn't respond, we will finally resort to calling emergency services (911) with the appropriate details to help the person who fell. This type of escalation procedure is something that sets us apart from other similar products that will automatically call 911, which would result in potentially unnecessary charges.
If the person who falls does not wish to reach out for help, that person will be able to prevent any calls/messages from being sent by simply pressing a button. This is a very intentional design choice, as there may be scenarios in which a person who falls is capable of getting up or the situation was somehow an accident. We will always resort to opting to help the faller, and simply allow that person to prevent notifications when necessary.
How We Built It:
We connect an MPU6050, an accelerometer and gyroscope, to an Arduino, which processes the raw data. We implemented an algorithm (C++) to use that raw data to predict whether a person has fallen or not with high confidence. We made sure to calibrate the algorithm in order to reduce the likelihood of false positives and false negatives.
The Arduino is connected to a Particle Photon, which is responsible for taking the prediction value from the Arduino and making an HTTP POST request (C++) to a REST endpoint that is built by using the StdLib-Twilio Hub (NodeJS). The logic within the StdLib-Twilio Hub is essentially our intelligent escalation notification system.
Finally, we took our device and created an accessible, user friendly wearable for any consumer.
Challenges We Ran Into:
None of our team have used the Photon Particle before, so we faced long challenges trying to understand problems with using the product and integrating it with our Arduino data. We also struggled with connecting our Arduino device to the Photon Particle, because there was not much documentation on the issues we faced.
What’s Next for FallCall:
While we spent much time tweaking our fall-detection algorithm, we can take it a step further and be able to use machine learning to more accurately customize a fall-detection algorithm based on physical feature data of a user. We would also love to improve the actual physical wearable, and make it more user-friendly to accommodate all potential users.