Project Resonance

Inspiration

Falls are the leading cause of injury-related death and nonfatal trauma among older adults as one in four adults aged 65+ experience falls annually. Existing wearable solutions detect falls reactively through impact alone, missing early-stage falls and generating false alarms from normal movement. We envisioned a smarter approach by combining kinematic and physiological data to distinguish genuine falls from everyday activities.

What it does

Project Resonance is a wrist-worn fall detector that fuses accelerometer and heart-rate data to identify critical falls in real-time. When a fall is detected, the device triggers haptic feedback (vibration) to alert the wearer and nearby caregivers. Crucially, it works entirely offline with no WiFi or internet required, ensuring reliability in areas with poor connectivity. Some key specs are a detection latency around 2 seconds and offline operation of 100%.

How we built it

Our hardware stack combines a HW123 gyroscope/accelerometer/temperature sensor for motion capture, a Gravity Heart Rate Sensor for biometric confirmation, and a Round 1.28" LCD display for real-time feedback to the wearer. For haptic feedback, we integrated the Titan Haptics Dev Kit alongside a piezo buzzer to ensure multiple alert modalities. We also included a speaker without driver for audio cues. The mechanical housing was custom 3D-printed (2–3 hours per iteration) to integrate all components into a wearable form factor. The software implements a two-stage fall detection algorithm running on the embedded processor. The first stage detects impact by monitoring the accelerometer output from the HW123 sensor when acceleration magnitude exceeds 5g, an impact event is triggered. The second stage confirms a genuine fall by cross-referencing the Gravity Heart Rate Sensor data, checking for a heart-rate elevation of > 20 bpm within 2 seconds of impact. This dual-stage approach effectively eliminates false positives from everyday movements like jumping or sitting down hard. When a fall is confirmed, the system activates haptic feedback through the Titan Haptics Dev Kit and the piezo buzzer while displaying an alert on the LCD screen.

Challenges we ran into

We faced several significant obstacles during development. The Gravity Heart Rate Sensor proved unreliable during motion, so we implemented a Kalman filter to smooth noisy readings and reduce false confirmations. Battery drain was another critical issue as initial firmware consumed too much power, limiting operational time. We solved this by switching to event-driven processing with interrupt-based wake-ups. Real-time processing on a low-power microcontroller presented latency constraints, but we optimized our algorithms using fixed-point arithmetic instead of floating-point calculations. Integrating multiple feedback modalities like the LCD display, Titan Haptics Dev Kit, piezo buzzer, and speaker required careful power management and signal timing. Finally, the 3D-printed enclosure needed mechanical isolation between the HW123 sensor and the haptic actuators to prevent vibration from the Titan Haptics Dev Kit from corrupting accelerometer readings; we solved this through careful housing design and post-processing filters.

Accomplishments that we're proud of

We designed a complete wearable system using accessible components (the Gravity Heart Rate Sensor, HW123 IMU, and Titan Haptics Dev Kit) while maintaining professional-grade performance. Our robust detection pipeline handles edge cases such as falls while exercising and sudden movements. We developed a multi-modal alert system combining haptic feedback, audio, and visual cues on the LCD display to ensure the wearer and nearby caregivers are immediately notified.

What we learned

Sensor fusion proved critical by combining the HW123 accelerometer/gyroscope data with the Gravity Heart Rate Sensor dramatically improved reliability. We discovered that optimizing code for microseconds truly matters on embedded systems. Integration complexity increases significantly with multiple feedback modalities; coordinating the Titan Haptics Dev Kit, piezo buzzer, LCD display, and speaker required careful synchronization. We found that 3D printing enabled rapid iteration on the mechanical enclosure, allowing us to test different sensor placements and isolation techniques. Test coverage was invaluable; we uncovered edge cases only through extensive real-world testing, not simulation. Finally, hardware-software co-design is essential as early choices about which components to integrate affected firmware complexity and power consumption.

What's next for Project Resonance

We plan to refine our machine learning capabilities by training neural network models on larger fall datasets to further reduce false positives. We are developing a smartphone companion app to enable caregiver notifications and data visualization. We are pursuing an FDA classification pathway to enable deployment in hospitals and care facilities as a regulated medical device. We aim to optimize our form factor by miniaturizing the system into a patch-based wearable using smaller sensors and haptic solutions, while maintaining the robust multi-modal feedback approach. Finally, we are exploring sensor expansion with additional environmental and biometric inputs to add contextual awareness to our fall detection pipeline.

Built With

• arduino
• c++
• gemini
• javascript
• node.js
• platform.io