Nobi recently secured a runner-up finish in the AI for Good Innovation Factory’s start-up competition. As a result, organizers for this month’s ITU’s AI for Good Global Summit 2024 asked Nobi to participate in the Summit because they believe Nobi is an inspiring example of how AI can make a tremendous impact on society and help advance the United Nations’ (U.N.) Sustainable Development Goals.
UN AI FOR GOOD GLOBAL SUMMIT
The Summit will be held in Geneva, Switzerland, May 30-31, with the goal of identifying practical applications of AI to advance the U.N.’s Sustainable Development Goals and scale those solutions for global impact.
A notable focus at the Summit will be demonstrations within healthcare, where AI can revolutionize different aspects of patient care and well-being, making it more personalized, efficient and accessible. Nobi will showcase how its AI-powered lamp for fall detection and prevention is revolutionizing elder care around the world.
As a lead-up to the Summit, we thought it would be a perfect time to talk to Stijn Verrept, Nobi’s co-founder and CTO, who is strategically involved in designing the AI capabilities for the company’s Smart Lamp.
He is passionate about creating solutions to help older adults age on their terms and enjoys the challenge of working at the intersection of care and technology. He’s seen how AI can address so many of the challenges associated with elder care, by both preventing and detecting falls which are the most significant cause of fatal injuries in older adults; taking the pressure off under-resourced caregivers in residential settings; and bringing peace of mind to families whose loved ones are living in their own homes.
When he speaks about the positive impact AI can have within healthcare sector around the world, you should listen. In Part 1 of this two-part blog series, Stijn talks about the AI capabilities of the Nobi lamp itself.
- Stijn Verrept, CTO en medeoprichter van Nobi
How does AI work in the Nobi Smart Lamp?
AI technology is embedded in Nobi’s Smart Lamp. The lamp includes an optical sensor that captures a room every second. Thanks to its powerful processor, it then determines various activities and objects within the environment. The AI has been trained on extensive data sets to recognize people, beds and different positions within a room. Consequences are linked to these observations, such as dimming the light when a resident goes to bed or illuminating the room when he or she gets out of the bed during the night.
Outside of fall detection, how else does Nobi utilize AI to support residents’ well-being?
Nobi adjusts lighting based on the time of day, promoting wakefulness during daylight hours and drowsiness at night. Nobi also provides detailed insights into residents’ sleep patterns, including when they go to bed, get out of bed and the total duration spent in bed. This helps regulate circadian rhythms and combat day-night rhythm distortion and confusion as well as mitigate risks associated with falls, especially for individuals in memory care. Additionally, alerts can be sent to caregivers when unusual behavior is detected, enabling timely intervention and support.
Facilities that have implemented Nobi lamps have seen the average response time of care staff to fall incidents dramatically decrease—to around two minutes. This significant reduction in response time enhances resident safety and well-being by ensuring timely assistance in the case of falls.
What distinguishes Nobi from its competitors in terms of AI?
Nobi stands out from its competitors by conducting all interpretations and analysis directly on the device, or lamp, itself—without the need to stream images to the cloud. This ensures privacy and security for users, as sensitive information remains processed locally.
What challenges has Nobi encountered in implementing AI for fall detection?
Nobi has faced numerous challenges, primarily due to the diverse environments in which it operates. One significant challenge was training the AI model to accurately detect falls without introducing bias. This required extensive training data, which was gathered through simulated falls and varied scenarios to ensure the model’s effectiveness across different situations.
How does Nobi address bias in AI training?
Bias in AI training is a critical concern for Nobi. To mitigate bias, the team employed various strategies, including wearing different clothing items, using props and capturing images in diverse environments. Additionally, Nobi continuously evaluates and refines the AI model through a closed-feedback loop, incorporating feedback from caregivers and residents to improve accuracy and reduce false positives.
How does the closed-feedback loop work?
Nobi utilizes a closed feedback loop to continuously improve its AI model. Every fall-detection event, whether accurate or a false positive, is evaluated and fed back into the system for further training. Caregivers provide input on detected events, helping to refine the model’s capabilities over time. This iterative process ensures that Nobi’s AI remains effective and adaptable to evolving needs and environments.
How does Nobi differentiate between actual falls and other similar scenarios, such as those involving pets or other objects?
Distinguishing between actual falls and other scenarios posed a significant challenge for Nobi. The team addressed this issue by incorporating diverse training data and refining the AI model to recognize specific characteristics associated with falls. Through continuous training and feedback, Nobi’s AI can now differentiate between human falls and other non-human activities with a high degree of accuracy.
What measures does Nobi take to ensure privacy and confidentiality in its AI implementation?
Nobi prioritizes privacy and confidentiality in its AI implementation. By processing data locally on the device, or lamp, itself, Nobi minimizes the need for transmitting sensitive information to external servers or the cloud. Any data that is transmitted is encrypted and anonymized, with personal identifiers removed to protect user privacy. Additionally, the company adheres to strict data protection protocols and obtains informed consent when necessary for training purposes, ensuring the privacy rights of residents are respected at all times.
In part two of this series, Stijn will discuss plans for future AI integrations in the Nobi lamp and how partnerships will play a key role in AI applications within the Nobi platform.