Fall Detector series

## Apple Watch Fall down series

showing it detected 12 out of 15 actual falls.

##A Low-Power Fall Detector Balancing Sensitivity and False Alarm Rate - IEEE

The Human Trials.
In the simulated fall (SF) trial, twenty young healthy volunteers (sixteen males and four females, height: 170.3 ± 8.3 cm, weight: 63.9 ± 11.0 kg, age: 27.6 ± 1.8 years) were recruited from staff and students at UNSW. The volunteers were asked to fall onto a foam mattress for safety, and the instructions of these simulated falls were designed based on real-world falls observed by a camera [16]. Before each fall, the
volunteers stood still on a platform at the same height as a foam mattress, and after each fall the volunteers lay still on the mattress for twenty seconds. Each simulated fall was repeated twice, totaling 200 fall events. A detailed breakdown of the protocol of simulated falls tested is shown in Table I.

In the free-living (FL) trial, ten young healthy volunteers (six males and four females, height: 167.8 ± 7.6 cm, weight: 62.6 ± 11.6 kg, age: 27.6 ± 1.5 years) were asked to wear the NEON-SD continuously for a day without any specific instructions. The volunteer could take the NEON-SD off if they felt uncomfortable while sleeping. It was verified that healthy young volunteers did not fall in the duration of the FL trial.

## Power-Efficient Interrupt-Driven Algorithms for Fall Detection and Classification of Activities of Daily Living - IEEE

A total number of 83 false positives were registered. This means that, on average, each WD had a false positive rate of 83/(24 × 3) ≈ 1.153 times per day.

Simulated Falls: 

As it is difficult to obtain real-world falls, tests of the proposed fall detection algorithm were simulated by four young subjects. Each of them wore a WD on his/her left wrist and performed 4 types of activities:

(1) walking, (2) walking upstairs, (3) walking downstairs, (4) walking some distance away, sitting down in a chair beside a desk and doing things such as writing, picking up and putting down random objects. Each person performed each activity five times for two minutes each time. Then activities (1), (2), (3) were performed again for five times, but each with

a fall in the middle. Test results are shown in Table I. It is observed that during continuous walking, falls were never be falsely detected as INACTIVITY interrupt did not have a chance to assert. While sitting down, wrists occasionally hit the desk with high impact, a false positive fall

will be accidentally triggered. It can also be observed that young subjects easily created false positives (3/20). In contrast, seniors in their real daily lives only creates 1.153 times per day as evaluated.

References: 

IEEE SENSORS JOURNAL, VOL. 15, NO. 3, MARCH 2015

Power-Efficient Interrupt-Driven Algorithms for Fall Detection and Classification of Activities of Daily Living

IEEE Journal of Biomedical and Health Informatics
A Low-Power Fall Detector Balancing Sensitivity and False Alarm Rate