Project Overview
Cox Communications partnered with Nottingham Spirk (NS) to create a wearable Mobile Personal Emergency Response System (mPERS) device designed to detect falls and minimize false alarms. The device included advanced sensors like accelerometers, gyroscopes, and pressure sensors. Plexar’s role was to develop a smart fall detection algorithm that could accurately identify real falls while avoiding false positives from everyday activities.
Despite significant progress in development and testing, the project was ultimately not released to market.
The Problem
Accurately detecting falls in a wearable device requires sophisticated algorithms and careful sensor integration to minimize false alarms and ensure reliable alerts. Publicly available datasets lacked the precision needed for real-world use, leading to unreliable results. The challenge was to build a custom algorithm that could handle a variety of fall scenarios while running efficiently on a wrist-wearable device with limited power.
The Solution
Plexar tackled the problem step by step:
- Custom Model Development: Plexar created a machine learning model tailored for fall detection, improving on the limitations of existing datasets.
- Real-World Data: Testing The team used real-time data from real-world scenarios to refine the algorithm, making it smarter and more accurate with every iteration.
- Prototype Creation: An initial prototype was built to test the algorithm in action, providing a foundation for ongoing improvements.
- Controlled Testing: To validate the algorithm, Plexar recreated testing conditions in India, simulating real-world scenarios to ensure reliable results.
- Final Algorithm Deployment: The fully developed algorithm was designed to run efficiently on the device, balancing accuracy, power usage, and flexibility with features like over-the-air updates for future improvements.
Results and Benefits
- High Accuracy: The final algorithm achieved over 90% sensitivity (accurate fall detection) and 95% specificity (non-fall activities).
- Power Efficiency: The algorithm was optimized to fit within the device’s power limits, ensuring long battery life without compromising performance.
- Thorough Validation: Testing under controlled conditions proved the algorithm’s reliability across different scenarios and environments.
- Future Flexibility: Over-the-air updates allowed for easy improvements to the algorithm without requiring major firmware changes.
Plexar played a key role in making the Cox wearable mPERS device smarter and more reliable. By building a custom fall detection algorithm and rigorously testing it, Plexar showed how technology can address complex problems in wearable safety systems.
Although the project didn’t reach the market, it highlighted Plexar’s ability to solve tough challenges and deliver innovative, real-world solutions for wearable devices.