🛡️ DoraShield — ELDORA Fall Detection Wearable

Protect. Respond. Recover.

ESP32 firmware for the ELDORA DoraShield — a wearable fall-detection device that continuously monitors acceleration, detects strong impact events, and triggers an immediate local buzzer alert to protect elderly users.

🌐 ELDORA Ecosystem · 🤖 DoraBot · 📱 ELDORA App

---

📌 Overview

DoraShield is the first line of defense in the ELDORA eldercare ecosystem. Worn on the body, it continuously reads inertial data from the MPU6050 accelerometer/gyroscope and triggers an immediate local buzzer alert the moment a fall or strong impact is detected — no internet connection required for the core safety response.

DoraShield's role in the ELDORA ecosystem: "The wearable guardian — always on, always measuring, always ready to raise the alarm before anyone even realizes something happened."

MCU	ESP32
IMU Sensor	MPU6050 (I2C) — 3-axis accelerometer + 3-axis gyroscope
Alert Output	Active buzzer (GPIO-driven)
Detection Method	Total acceleration vector magnitude threshold
Power	Li-ion battery + power module
Response Latency	Real-time (continuous polling loop)

---

🌐 ELDORA Ecosystem

DoraShield is the Protect layer of ELDORA's three-phase safety framework:

ELDORA Ecosystem
├── 🛡️  DoraShield   — Fall detection wearable (this repo — ESP32)
├── 🤖  DoraBot       — AI voice companion (ESP32-S3)
└── 📱  ELDORA App    — Caregiver dashboard (XGBoost + SHAP, Isolation Forest)

---

✨ Firmware Features

• 📡 MPU6050 Initialization — configures the IMU over I2C on boot with error checking
• 🔄 Continuous Acceleration Reading — polls 3-axis accelerometer data in the main loop
• 📐 Total Acceleration Vector Calculation — computes resultant magnitude from X, Y, Z axes
• 💥 Fall / Impact Detection — triggers alert when magnitude exceeds the configured threshold
• 🔔 Local Buzzer Alert — activates buzzer immediately on detection, deactivates when motion normalizes
• 🖥️ Serial Monitor Output — streams live sensor values and alarm state for debugging and threshold tuning

---

🛠️ Tech Stack

Layer	Technology	Purpose
MCU	ESP32	Main processor, I2C master, GPIO control
Framework	Arduino (ESP-IDF)	Firmware development environment
IMU Driver	Adafruit MPU6050	I2C communication and sensor abstraction
Sensor Abstraction	Adafruit Unified Sensor	Standardized sensor event interface
I2C Bus	Wire	Arduino I2C protocol library
Debug Output	Serial Monitor (115200 baud)	Live sensor data + alarm state logging

---

🔧 Hardware

Component	Model / Spec	Role
Microcontroller	ESP32	Main firmware host, I2C master
IMU	MPU6050	3-axis accelerometer + 3-axis gyroscope
Alert	Active buzzer	Immediate local fall alarm output
Power	Li-ion battery + power module	Portable wearable power supply

Pin Configuration

Defined in FallDetection.ino:

Pin	GPIO	Connected To
I2C SDA	GPIO 4	MPU6050 SDA
I2C SCL	GPIO 5	MPU6050 SCL
Buzzer	GPIO 3	Active buzzer signal

⚠️ If you rewire the hardware, update the #define constants at the top of FallDetection.ino to match your new pin assignments before flashing.

---

📁 Project Structure

DoraShield/
│
└── 📄 FallDetection.ino     # Complete firmware — MPU6050 init, read loop,
                             # vector calculation, threshold check, buzzer control

Why it's a single file

DoraShield's firmware is intentionally minimal and self-contained. The entire detection pipeline — sensor init, data read, math, alert output, serial logging — fits cleanly in one Arduino sketch, making it straightforward to flash, audit, and tune. There are no hidden dependencies between files or initialization order surprises.

---

⚙️ How DoraShield Works

In plain English:

DoraShield boots → initializes MPU6050 over I2C
    → enters continuous loop:
        reads ax, ay, az from MPU6050
        computes total acceleration magnitude
        if magnitude > threshold → buzzer ON  (fall detected)
        else                     → buzzer OFF (normal motion)
        streams readings to Serial Monitor

---

🔬 Fall Detection Logic

The core detection is a total acceleration vector magnitude check:

// Read accelerometer, gyroscope, and temperature events
sensors_event_t a, g, temp;
mpu.getEvent(&a, &g, &temp);

// Compute resultant acceleration magnitude across all three axes
float totalAccel = sqrt(
  a.acceleration.x * a.acceleration.x +
  a.acceleration.y * a.acceleration.y +
  a.acceleration.z * a.acceleration.z
);

// Threshold-based fall detection
if (totalAccel > 15.0) {
  digitalWrite(BUZZER_PIN, HIGH);  // FALL / IMPACT detected
} else {
  digitalWrite(BUZZER_PIN, LOW);   // Normal motion
}

Why total magnitude?

A fall produces a characteristic impact spike — a sudden surge in resultant acceleration across all axes simultaneously — regardless of the wearable's orientation. Using the vector magnitude instead of individual axes makes the detector orientation-agnostic, so it works whether DoraShield is worn on the wrist, chest, or clipped to a belt.

Tuning the threshold

The default threshold of 15.0 m/s² is a starting point. To tune it for real-world accuracy:

Step	Action
1	Open Serial Monitor at 115200 baud
2	Have the wearer perform normal activities (walking, sitting, standing up)
3	Note the peak totalAccel values during normal motion
4	Simulate a controlled fall onto a soft surface and note the spike value
5	Set threshold between the highest normal-motion peak and the lowest fall spike

General guidance:

Scenario	Typical totalAccel
Standing still	~9.8 m/s² (gravity only)
Walking	10–13 m/s²
Sitting down quickly	12–14 m/s²
Fall / hard impact	> 15–25 m/s²

A threshold too low → false positives on vigorous movement. Too high → missed falls. Tune per user and wearable placement.

Extending the detection algorithm

The current firmware uses a simple single-threshold check. More robust approaches to consider:

Enhancement	Description
Free-fall pre-phase	Detect a brief near-zero-g window (< ~2 m/s²) before the impact spike — characteristic of true falls
Gyroscope fusion	Add angular velocity check: falls often produce rapid rotation on at least one axis
Debounce window	Require the threshold to be exceeded for N consecutive samples before triggering
TinyLSTM (future)	Replace threshold logic with a trained time-series classifier for higher precision

The ELDORA roadmap targets a TinyLSTM model on ESP32 as the next detection layer for reduced false positives.

---

⚙️ Configuration

Update the following constants in FallDetection.ino before flashing:

// ── Hardware Pins ─────────────────────────────────────────
#define I2C_SDA    4   // MPU6050 I2C data
#define I2C_SCL    5   // MPU6050 I2C clock
#define BUZZER_PIN 3   // Active buzzer output

The fall detection threshold (15.0 m/s²) and serial baud rate (115200) are set inline in the loop() and setup() functions respectively. Update them directly in FallDetection.ino to suit your hardware and deployment environment.

---

🚀 Build & Flash

Prerequisites

• Arduino IDE 2.x with ESP32 board support installed
• ESP32 board package URL: https://raw.githubusercontent.com/espressif/arduino-esp32/gh-pages/package_esp32_index.json

Required Libraries

Install via Arduino Library Manager (Sketch → Include Library → Manage Libraries):

Library	Install Name	Purpose
Wire	(bundled with ESP32 core)	I2C bus communication
Adafruit MPU6050	Adafruit MPU6050	IMU driver and I2C abstraction
Adafruit Unified Sensor	Adafruit Unified Sensor	Standardized sensor event interface

ℹ️ The Adafruit MPU6050 library depends on Adafruit Unified Sensor — the Library Manager will prompt you to install both.

Flash Steps

# 1. Clone the repository
git clone https://github.com/Eldoraaa/dorashield.git
cd dorashield

# 2. Open FallDetection.ino in Arduino IDE

# 3. Select board:
#    Tools → Board → ESP32 Arduino → ESP32 Dev Module
#    (or the specific ESP32 variant you are using)

# 4. Update pin and threshold constants in FallDetection.ino

# 5. Connect ESP32 via USB, select the correct COM port

# 6. Upload
#    Sketch → Upload  (or Ctrl+U)

# 7. Open Serial Monitor at 115200 baud to observe live sensor output

---

👥 Team

ELDORA — BINUS BM Team Passage to ASEAN Hackathon 2026

Name	Role
Stanley Nathanael Wijaya	Team Lead
Lutfi Alvaro Pratama	IoT Engineer
Andrian Pratama	Mobile Developer
Khalisa Amanda Sifa Ghaizani	Backend Developer
Devon Nicholas	AI Engineer

---

📧 Contact

Have questions, want to collaborate, or interested in ELDORA?

Channel	Details
📧 Email	stanley.n.wijaya7@gmail.com
✈️ Telegram	@xstynwx
💬 Discord	stynw7

---

Made with 🤍 by BINUS BM Team 🔥