Process

The first version of the device detected the position of a person. We found that for most of the daily movement of a person, the upper body of a person maintains an upright posture. But in the case of a fall, a person will lie on the ground eventually. This result in the posture change of the upper body. Our first algorithm was based on it. 

The second version of the device deteced impact, in other words, a sudden increae of acceleration. This acceleration can be caused by fall as well as some daily movement. Hence, an algorithm to eliminate such noises are necessary. However, we found out that the movement of a human body was quite complex, and it was not just a matter of sensor sensitivity.

The third version of the device detected both impact and position change. We did many user tests to revise the code. These tests include walking, bending, going up/down stairs and so on.  The final version of the fall detector passed all these activities. However, there was still some noises which could not be ultimately eliminated.