Inspiration
We were inspired by the desire to provide a simple yet effective solution to help protect seniors in their everyday lives. Many elderly individuals live alone or spend long periods of time without direct supervision, making fall detection and immediate assistance crucial for their wellbeing. We wanted to create something that offers peace of mind to both seniors and their loved ones.
What it does
HaloGuard is a lightweight computer vision application that uses a standard camera to detect potential falls. It monitors the user’s posture, flags prolonged “fallen” states, and triggers an alert if it detects a high likelihood that someone has fallen. It can also detect if the user is waving for help, ensuring both passive and active monitoring for safety events.
How we built it
• Python for the main application logic
• OpenCV for webcam access and image manipulation
• Mediapipe for real-time human pose detection
• Custom fall detection logic to watch for a user’s orientation and ensure the alert only triggers if the fallen state persists for a few seconds
• Threading to handle audio alerts in the background, so the program’s main loop remains responsive
Challenges we ran into
• False Positives: Distinguishing between someone actually falling and simply bending down or sitting on the floor required careful thresholding and time-based checks.
• Camera Placement: Ensuring the system worked in real-world conditions (different room layouts, lighting, or angles) was tricky.
• Balancing Sensitivity: We wanted the app to be sensitive enough to detect true falls but not over-alert for everyday motions.
Accomplishments that we're proud of
• Consistent Fall Detection: We fine-tuned the logic so that brief changes in posture don’t trigger false alerts.
• Waving Detection: We successfully incorporated a simple gesture detection so users can actively signal for help.
• User-Friendly Design: We minimized configuration steps, aiming to make it accessible for non-technical users.
What we learned
• Practical Computer Vision: We gained deeper insight into real-time pose estimation and how to filter out noise.
• Iterative Thresholding: Fine-tuning detection thresholds requires testing with multiple people and real scenarios.
• Threading & Performance: We learned how to manage tasks in parallel without blocking the main video loop.
What's next for HaloGuard
• Mobile Integration: Potentially porting the core logic to a mobile app for greater portability and accessibility.
• Smart Alerts: Sending notifications to family members or caregivers through SMS or an app.
• Additional Sensors: Integrating optional environmental sensors (like motion or pressure sensors) to bolster reliability.
• Refined Interface: Introducing a simple dashboard or voice prompts to guide setup and confirm alerts more intuitively.
Log in or sign up for Devpost to join the conversation.