IoT Fall Detection System for Elderly

Elderly people living alone face a quiet, serious risk: a fall nobody witnesses. A hip fracture or head injury can leave a person unable to reach a phone, and the hours spent waiting for help often decide how well they recover. Families worry, but round-the-clock supervision is neither affordable nor welcome — nobody wants a camera in a parent's bedroom. Commercial fall detectors exist, yet they are expensive, subscription-locked and opaque about how detection works. A student-built system must solve the same hard problems honestly: telling a real fall apart from sitting down hard, dropping the device or bending over, and raising an alert without depending on home WiFi. This project does that with a wrist-worn ESP32 and MPU6050 IMU, implementing the classic three-stage fall signature — free-fall dip, impact spike, post-fall inactivity — plus a cancel button so false alarms never needlessly panic the family.

1. The MPU6050 continuously samples acceleration and angular velocity at 100 Hz; the ESP32 computes the resultant acceleration vector magnitude.
2. A fall signature is flagged when the algorithm sees the three stages in sequence: a sudden free-fall dip (resultant < 0.6 g), a sharp impact spike (> 2.5 g) within ~1 second, then an inactivity window (minimal movement for ~10 seconds).
3. On detection, the buzzer sounds and a 20-second cancel countdown starts on the device — if the wearer is fine, they press the red cancel button and nothing is sent.
4. If the countdown expires, the GSM module (SIM800L) texts all registered emergency contacts with the fall time and a GPS location link.
5. The event is logged with its timestamp and sensor trace for later review in the report.

• MPU6050 6-axis IMU (3-axis accelerometer + 3-axis gyroscope) in a wrist/belt-worn enclosure
• Three-stage fall signature: free-fall dip → impact spike → post-fall inactivity window
• Configurable sensitivity thresholds tuned for elderly movement profiles
• 20-second cancel countdown with buzzer — one button press stops false alarms
• GSM SMS alert to up to 5 emergency contacts with GPS location link
• On-device buzzer pre-alert sounds before the SMS goes out
• NEO-6M GPS module embeds a Google Maps link in the alert SMS
• Low-power design — ESP32 deep-sleep with motion-interrupt wake
• Activity log: fall events with timestamps stored on-device and viewable in the app
• Alert app view: fall-event notification card, location map and emergency contact list

• Fully assembled fall-detector wearable
• Complete firmware source code (Arduino IDE, well-commented)
• Circuit diagram and wiring schematic
• Fall-algorithm tuning guide (threshold explanation for the viva)
• Project report PDF (synopsis, design, test procedure, conclusion)
• PPT presentation for final review
• Viva Q&A preparation document
• GSM SIM setup and contact-registration guide

FAQs

1. How accurate is the fall detection? The three-stage signature (free-fall → impact → inactivity) rejects most everyday motions like sitting or bending. Design target is ~90%+ sensitivity on simulated falls once thresholds are tuned; the report documents the evaluation procedure and the false-alarm cases to watch for.
2. What stops false alarms from alerting my family? The 20-second buzzer countdown. Every detection first warns the wearer; a single press of the red button cancels the alert. The SMS only goes out if nobody cancels.
3. Does it need the internet? No — alerts go over the GSM cellular network as SMS, so it works anywhere with mobile signal, even without WiFi or a paired smartphone.
4. How long does the battery last? The ESP32 sleeps in deep-sleep mode and wakes on motion interrupts, giving roughly 2–3 days per charge on an 18650 cell. The report includes the full power-budget table.
5. Can I demo this safely? Yes — drop the device onto a cushion or simulate the motion profile; the buzzer countdown and (test-mode) SMS fire exactly as in a real fall. A safe demo script is included.
6. How can I extend this project? Common extensions include: heart-rate (MAX30102) anomaly fusion, LoRa instead of GSM for campus-scale coverage, fall-direction classification, and a caregiver cloud dashboard with history.

• This is an engineering prototype, not a certified medical device — a fall alert aid, not a diagnostic tool.
• GPS needs a clear view of the sky; indoors, the alert goes out without a location link.
• A working SIM card with SMS balance and network coverage is required for alerts.
• Vigorous activities (dropping the device, sudden sitting) can trigger false detections — the cancel button is the designed safeguard.