Fall detection for an independent and self-determined life
The autonomy of intelligent machines is continuously increasing, and with it, the ability to take over tasks originally reserved exclusively for humans. The opportunities this presents for demographic change and an aging society are immense. New technologies aim to simplify the lives of older people and promote independence into old age. At the same time, the care burden can be reduced for both families and care facilities. The use of artificial intelligence, in its current form of machine learning, is undoubtedly one of the key technologies and helps to interpret human movement.
Fall detection is an important technology in the field of assistance systems for older people and people with disabilities. With the help of sensors and algorithms, a fall can be detected and an emergency call can be triggered automatically. This allows affected individuals to live more independently and autonomously, while increasing safety in the event of a fall.
Falls are common occurrences in older age groups. One in three people over 65 falls once or several times a year¹. More than half of those who fall are unable to get up independently due to age-related physical limitations. It can take hours or even days for help to arrive².
For older people, falls usually have far-reaching consequences. They often lead to fractures, loss of mobility, limitations in independence, need for care, or even death. Furthermore, they can have a negative impact on self-esteem, activity, and the social contacts of those affected².
Expert Interview: Challenges and Opportunities of Artificial Intelligence in Fall Detection
Today, there are numerous technical solutions that detect falls in real-time and quickly and easily notify relatives or professional emergency call services. We asked our Data Scientist, Jan Peukert, who is working on the development of our fall detection, for his insights.
Jan Peukert focuses on recording, analyzing, and evaluating sensor data and develops various algorithms that are integrated into the Gardia emergency bracelet.
Mr. Peukert, what technical solutions does the market offer in the field of fall detection?
Probably the best-known solution on the market are fall detectors. These sometimes only detect hard falls and require the affected person to remain motionless for up to 30 seconds after a fall in some cases, only then triggering an automatic emergency call. However, in emergency situations, it is often observed that fallen persons can still move or twitch an arm. In most cases, this does not trigger an automatic emergency call.
In this case, the affected person must request help via an emergency button. If the fallen person is unable to press the emergency button, they may remain on the floor for hours, waiting for timely assistance. As you can see, this solution has serious disadvantages.
What distinguishes your fall detection?
Gardia's fall detection is based on data from three different sensors and attempts to classify events into everyday events and fall events. The interaction of these three sensors allows us to achieve a more robust and highly accurate fall detection. We have also optimized the analysis of the sensor data to such an extent that we can achieve a battery life of up to 21 days. We achieve this by offloading the sensor data analysis. The built-in "wake-up function" activates the sensor only for critical events and leads to an appropriate analysis. We determine various features that are then forwarded to neural networks, which ultimately make the final decision as to whether an event is a fall or not.
For training the neural networks, various types of falls, such as "falling while walking" or "tipping from a chair", were simulated. Everyday events were not simulated in the lab, but recorded by various people in their daily lives to obtain the most realistic data possible.
Often, specific events are simulated, but it is unclear whether these events are critical, how often these events occur in a critical form in everyday life, and whether the events are realistically simulated in the lab. By recording entire days over several weeks, the neural networks can be trained with realistic data, leading to intelligent fall detection that correctly identifies almost all falls.
Artificial Intelligence for Fall Detection: a Promising Solution in Medical Care
The use of artificial intelligence for fall detection is one of the most important applications in healthcare. With the help of motion capture and analysis, falls can be automatically detected and appropriate measures can be taken to prevent further health consequences in addition to the injuries sustained by the fallen person, or to provide quick assistance.
Fall detection using artificial intelligence is an important and promising area of medical care. It offers the opportunity to improve the lives of older people and people with physical limitations, while simultaneously reducing the care burden.
And here's how Gardia's intelligent fall detection works:
