OUTPUT
Test using walking data
The function used to calculate r square value
The testing of walking data with the program returned all extremely negative values for the sum of r-squared for each axis, which determined that this was not a fall.
The result indicating that walking is not a fall
Test using fall data
The validation for unknown falling data had returned a positive value for Falling Left and Falling back. However, the acceptable value was only found with Falling Left, which was then confirmed that the result is true with the original test data on which side they had fallen on. This confirmed the quality of the program with a moderate accuracy for fall detection, as well as which side they fell to.
The result indicating that walking is not a fall
Value
Creation
Our project idea is to create a system that would alert emergency contacts when an end-user has taken a fall. Our end-users are people who are vulnerable to falling, including patients with walking disabilities, and elders. It seems that for this project, the most accessible and commonplace sensor that we are able to acquire and utilize is the smartphone, which has a 3-dimensional accelerometer, gyroscope, and other various sensors such as the altitude sensor or GPS and for utilization. The fall detecting system has a tremendous value for our end-users since timely medical treatment can save their lives. These people can end up in a situation where they are unable to receive help immediately - which is a problem that this system could help eradicate.
​
The created proof of concept can be utilized by any type of accelerometer but is most probable to use a phone’s accelerometer. This is because, in the 21st century, the one thing that people carry around all the time is their smartphone - it has the computational power to run these algorithms, and it is on them all the time - which is perfect for continuous data monitoring and detection. With the app running in the background, a person carrying a smartphone would be able to fall and get help in a situation where they would not be able to get back up by themselves, utilizing the accelerometer on the phone as well as other tools such as the GPS or messages to deliver the crisis at hand.
Future
extensions
Further research can be taken to various extents. We believe that for an emergency situation to be detected on the phone, there can be three cases to explore. First, a steep decrease in altitude can be detected on an altimeter implies a sudden fall from a high altitude area to a low altitude area. Second, an abrupt change in the motion of the accelerometer can mean various things - we believe that there is a method to characterize the shape of the signals in the accelerometer when a person has fallen. Third, if a person does not move at a remote location, detected by a GPS for a period of time, there is a chance of trauma that the end-user cannot respond to. Of course, this case would have to take inputs as a parallel with other forms of fall/emergency detection.