No, a loss of verticality is not necessarily related to a fall. This is especially true when it comes to an alert device in the form of a mobile lone worker application, where the smartphone can be used for other purposes.
It’s important to note that falls accounted for 15% of workplace accidents and 21% of workplace accident-related deaths, according to a 2022 report by CNESST. Therefore, it is the primary risk of accidents and deaths in the workplace, as identified by a QHS manager through a comprehensive risk analysis.
As a QHS manager, you may have already started looking for an alert solution to implement in your company as part of your commitment to safety. However, despite your best intentions, there are many players in the field, and they all seem to offer similar features: fall detection, loss of verticality, SOS, and immobility, … You may find yourself a bit overwhelmed in the jungle of alert devices, making it difficult to differentiate between them.
But don’t worry! By the end of this article, you should have a clearer understanding. Let’s demystify one of the crucial features of any lone worker protection solution: fall detection.
The goal of this article is to provide you with enough knowledge to distinguish between an innovative solution and one that just plays with words! This way, you can choose an effective lone worker protection system that meets your needs.
The ultimate goal of any alert application is to enable the timely response of emergency services when an accident occurs to a lone worker. Initially, the first alert applications simply provided an alarm button, allowing lone workers to manually trigger an alert in case of an accident.
With the development of electronics in the 1980s, the first automatic fall detection functions emerged. These devices, often referred to as « Man-down » devices, simply detect prolonged immobility of the lone worker and automatically trigger an alert after a certain period has elapsed.
The operation is as follows:
Hence the use of the term « Man-down device” (referring to a person who is no longer moving).
MUnfortunately, the issue caused by this type of device is double:
Despite these limitations, there are still products available on the market that offer this functionality.
To address the drawbacks of fall detection based on immobility detection, the range of alert applications has evolved by introducing a new type of detection: loss of verticality.
In the context of the loss of verticality detection, the alert application constantly monitors its angle of inclination concerning a vertical axis. As soon as the smartphone inclines to a specific angle, often arbitrarily determined by the manufacturer or operator, the system triggers the alert.
The operation is as follows:
This represents a significant shift in approach compared to the previous method. The high-risk worker can now work while being stationary (sitting for all or part of the day) without triggering false alarms that would disrupt his work.
However, this introduces a new issue:
While a fall necessarily leads to the loss of verticality, the reverse is not necessarily true!
A mechanic lying down under a vehicle or a logistics agent bending to pick up a box does not necessarily mean that they have experienced a fall… However, the alert application still triggers the loss of verticality alarm.
Therefore, fall detection through loss of verticality requires you to constantly carry your smartphone vertically, and avoid horizontal work. The detection of loss of verticality can be relevant for certain activities but counterproductive for many others.
Maintaining the smartphone in a vertical position can be quite challenging, leading to a significant number of false alarms even when the worker is still standing. This can greatly disrupt their work.
This third evolution is more recent, and few alert application providers offer it today. The reason is simple: detection of a fall is an extremely complex phenomenon to determine.
Indeed, it is easy for you to see if a person has fallen when they are on the ground. However, close your eyes and try to exhaustively describe all the conditions that can help determine whether or not a fall has occurred. The exercise becomes more complicated, doesn’t it? Now, imagine the complexity involved in configuring an alert application to detect a fall!
Visual analysis of a fall: (Fall image) This is a fall easily recognizable through visual observation.
Descriptive analysis of a fall: A fall can be broken down into three stages:
This approach raises 3 questions:
Manufacturers who offer this functionality each try, in their way, to construct an algorithm, defining various parameters (angle of inclination, impact, immobility…) to varying degrees of effectiveness.
The method of fall detection through fixes threshold involves precisely defining the criteria that qualify a fall. The problem is that there are many types of falls (fall from a ladder, fall down stairs, running fall, soft fall…). The task is therefore very complex, and this method quickly shows its limitations. Furthermore, it has the major disadvantage of providing a false sense of confidence. You may test fall detection in a very specific case and be reassured because you obtain satisfactory results. However, in other situations, your alert application may not detect the fall because it does not meet the arbitrary criteria set by the manufacturer to qualify as a fall! The use of a different smartphone can also impact the result.
Faced with the limitations of fall detection through fixes threshold, some manufacturers have chosen to collaborate with specialized AI scientists (Data science experts) to work on this issue.
Here, the approach is radically different. The phenomenon of falling is no longer defined by words but by the data collected from users. The challenge of this approach lies in the significant amount of data that needs to be collected for AI to distinguish what constitutes a fall from what does not. Once this hurdle is overcome, the results of this method are incomparable to the fixes threshold approach. Furthermore, it has the huge advantage of being less dependent on the quality of the smartphone’s sensors.
You will have understood that this fourth evolution shifts the alert application into the world of AI and continues to revolutionize fall detection.
The scientific community regularly publishes advancements in this field.
Through this article, we have identified four ways to detect the fall of a lone worker:
While loss of verticality is not recommended today, immobility detection may have its merits but only as a complement to true fall detection. For example, in the case of a person having a fainting fit while sitting at their desk. Therefore, immobility detection should be considered as a supplementary function to reduce the risk of missing an accident.
As for the fixes threshold fall detection, it should be approached with caution, and true AI-based fall detection should be preferred over it.
Our recommendations are as follows:
Recommendation number 1: Analyze your constraints
Clearly define your needs and the working constraints of your employees. Do they work lying down? Are they sometimes immobile in their tasks?...
Recommendation number 2: Analyze your equipment
What sensors are required on the smartphone to detect falls (accelerometer, gyroscope, barometer…)? Does your current fleet of professional smartphones have these sensors?
Recommendation number 3: Test in real conditions before committing
This recommendation is probably the most important because it allows you to verify that the announced fall detection is of good quality and not disguised loss of verticality detection. Some manufacturers may use misleading terminology and classify a simple loss of verticality detection as fall detection. Others may claim to have AI-based fall detection without having collected sufficient data for true fall detection.
This testing approach takes more time at the outset, but in the long term, it will be beneficial! It will allow you to achieve a real return on investment without the need to change providers at the end of each engagement.
A confident provider in its solution will never refuse a test! In the realm of alert applications, the cheapest solution often ends up costing you the most over time. Constant false detections will test the patience of your lone workers, and poor detections will erode their trust.
If you are looking to detect real falls, test our VigieApp application, which, thanks to an innovative AI algorithm, can differentiate between falls detected through the fixes threshold and loss of verticality.
