The results show the overall accuracy of 84.72%, and
overall false positive during ADL (false positive = the device
captures a fall but it is a wrong detection) as 1.94%. Although
the experiments revealed impressively accurate results, some of
the participants’ fall detection accuracy is lower than expected.
Possible reasons for these inaccurate detections are as follows:
It is difficult to record all possible daily behaviors (ADL) in a
limited time period. Another issue is that the participants were
anxious about a real fall, which can cause serious physical
damage, and therefore, the deliberate falls that were recorded
were unnatural, resulting in a lower fall detection accuracy.