ManDown Detection
When it comes to ManDown recognition, we are faced with a wide range of applications and use cases. Identifying a dangerous condition for the person on the ground may not be easy, as we could find ourselves faced with various situations (all equally critical) but which present completely different evolutions. Just think of the difference between a person who slowly collapses on the ground (bending over and then going to the ground), compared to a person who is hit by a vehicle or who falls abruptly on the ground.
Being able to have a single device that is able to identify such different situations at the same time is a difficult task, therefore, usually a use case is preferred and the best operating parameters for that type of application are selected.
For what it concerns BlueUp devices, when properly configured, some BlueUp firmwares have the capability to detect ManDown events. A ManDown event can be described as a human person fall followed by a time interval in which the fallen person remains still on the ground. ManDown detection is done by using the tag integrated accelerometer sensor in conjunction with a proprietary algorithm developed by BlueUp.
The accelerometer sensor is used to simultaneously detect a fall event and to monitor the motion state of the person wearing the tag.
From the tests conducted in laboratory it emerged that a ManDown event can be split into 3 sections:
- the fall event
- the ground impact, with possible rebound of the body settling on the ground
- the still body lying on the ground
In the below image is shown a plot of the three accelerometer axis recorded during a test fall with the 3 ManDown algorithm stages highlighted.
Based on the previous considerations, the accelerometer must be able to:
- identify the fall event
- monitor the motion state
These features must be performed simultaneously.
The fall event recognition can be done in 3 (independent) ways:
- Position Detection : the accelerometer monitors the position of the tag, the fall event is detected everytime the tag is in horizontal position
- Freefall Detection : the accelerometer monitors the acceleration of the tag, the fall event is detected everytime the tag acceleration on the 3 axis is around 0 (that is the freefall condition)
- Shock Detection: the accelerometer monitors the acceleration of the tag, the fall event is detected everytime the tag observe an instant acceleration above a certain threshold.
Once the fall event is detected (using one of the above criteria), the accelerometer will continue to observe the motion state of the tag for a configurable time interval, called Debounce Duration. This time interval is time that the accelerometer waits for the tag to stop bouncing on the ground (due to the ground impact).
Once the bouncing is over, the accelerometer actively monitors the tag moving state, for a configurable interval called Stationary Duration. At this stage, if the accelerometer detects amovement, it assume that the fall event was a false positive (because the body is moving after the fall event and the debounce interval) and abort/reset the ManDown detection algorithm. If instead, no motion is detected during the whole Stationary time interval, the accelerometer will signal the ManDown event to the application.
Accelerometer Configuration
The very first parameter to setup when configuring a tag for ManDown recognition is the fall detection mode. As stated above, fall detection in ManDown algorithm can be configured with different criteria, that are: Position Detection, Freefall Detection and Shock Detection.
Position Detection
In this mode the accelerometer is used to monitor the tag position: horizontal or vertical.
This mode of operation requires the configuration of the following parameters:
- Angle (deg), define the angle threshold that is used to switch between horizontal and vertical position, range [1, 90]
- Duration (sec), define the time duration in which the position should be kept stable to trigger a horizontal/vertical transition, range [1, 255]
When this mode is coupled with the ManDown option, the fall event is associated to the horizontal position of the tag.
Since the tag, if not be properly hooked with the person body, could be in a vertical position after the fall, this accelerometer mode is not recommended for real ManDown use cases.
FreeFall Detection
In this mode the accelerometer is used to recognize freefall events.
From the accelerometer point of view, a freefall event is detected when all the three axes (x, y, z) approach to 0g acceleration simultaneously.
This mode of operation requires the configuration of the following parameters:
- Freefall height (cm), define the minimum height of the freefall (every freefall from a height greater or equal to this value will be recognized), range [50, 255]
- Freefall sensitivity, define the sensitivity of the freefall detection, a small value will detect only perfect freefall (where acceleration really approaches to 0g) while a high value will be more tolerant (acceleration below t = sensitivity · 50mg), range [1, 10]
The usage of this accelerometer mode for real ManDown recognition is not recommended, since the three axes never approach 0g acceleration simultaneously when a person falls from the stand up position to the ground.
Shock Detection
In this mode the accelerometer is used to detect an unexpected strong instant acceleration of the tag.
This mode of operation requires the configuration of the following parameters:
- Shock intensity, define the intensity of the shock to recognize, that means that the accelerometer will detect events that cause an instant acceleration above the threshold t = 1500 + intensity · 50mg
This mode is the BlueUp recommended one since each fall will produce a shock event that can be detected from the accelerometer.
The other configuration parameters for ManDown detection are:
- Debounce Duration (sec), is the maximum time interval in which the body of the fallen person is bouncing when impacted on the ground, range [0, 255].
- Stationary Duration (sec), is the minimum time interval in which the person remains stationary on the ground after falling, range [0, 255].
- Alert Duration (sec), defines the time duration, after ManDown has been detected, during which the tag emits an alert sound and after which the tag starts sending the ManDown alarm event through the network, range [0, 255].
If the Alert Duration is equal to 0 the tag does not perform any alert sound after ManDown detection and will immediately send the ManDown alarm event through the network. If the Alert Duration is greater than 0, the tag will emit a sound alert for Alert Duration seconds before sending the data through the network.
When the Alert Duration is greater than 0 and the tag starts emitting the alert sound, the user can choose to abort the ManDown detection if it was a false positive case, by pressing the tag button. In this case the ManDown alarm event will not be sent through the network.
Test report
The ManDown algorithm was tested by BlueUp.
Choosing the right shock threshold value is a compromise:
- If we choose a too small value, the accelerometer shock detection will be too sensitive and every small shock event will be detected from the accelerometer. This will allow us to be sure that when a fall event occurs, the shock will be detected for sure, but also other small shock events will also be detected during normal daytime activities like walking or sitting down and this could produce false positives.
- If we choose a too high value, the accelerometer shock detection will be less sensitive and only high shocks will be detected from the accelerometer. Small shock events of daytime activities won’t be detected, but if the value is too high, even a fall could be missed.
After some tests performed, in the following is reported the suggested configuration that provided best results for real ManDown detection:
- Motion Threshold = 10 (160mg)
- Motion Duration = 2 sec
- Fall detection mode: SHOCK
- Shock Intensity = 2 (1600mg)
- Debounce Duration ≥ 3 sec
- Stationary Duration ≥ 10 sec
- Alert Duration = 60 sec
With such configuration a 90% success rate has been obtained.
The above success rate has been obtained using the above configuration parameters, with the tag attached with the clip to the pocket or to the belst (to prevent tag swinging that could interfere with the correct functioning of the ManDown algorithm).
Other wearing modes have not been tested. Please contact BlueUp for suggestions.
The supplied device is not to be considered a safety device. BlueUp indemnifies itself from any responsibility for any type of damage or accident for which the device has not reported the alarm event
Il dispositivo fornito non è da conisiderare un dispositvo di sicurezza. BlueUp si indenne da ogni responsabilità per ogni tipo di danno o incidente per il quale il dispositivo non abbia segnalato l'evento di allarme