Fog-Based Real-Time Sleep Quality Monitoring System

A Cyber-Physical System that monitors sleep quality in real-time using wearable sensors, Fog Computing, and a TensorFlow Lite LSTM model — all processed locally with zero cloud dependency.

---

Architecture

┌──────────────────────────────────────────────────────┐
│                    FOG LAYER (Laptop)                 │
│                                                      │
│  Preprocessing  →  TFLite LSTM  →  Streamlit Dashboard│
│  (Feature Eng.)    (64→32 units)    (Live Score + HR) │
└──────────────────────────────────────────────────────┘
                        ↑
              USB Serial (115200 baud)
                        ↑
┌──────────────────────────────────────────────────────┐
│                 EDGE LAYER (Arduino)                  │
│                                                      │
│  Arduino UNO + MPU-6050 (Accel) + PPG Pulse Sensor   │
│  Output: timestamp, AcX, AcY, AcZ, Pulse @ 10Hz     │
└──────────────────────────────────────────────────────┘

Flow: Arduino reads sensors → sends CSV over serial → Laptop (Fog Node) runs TFLite inference → Streamlit shows live results.

---

Dataset

MMASH (PhysioNet) — 22 real subjects, 1.4M data points of accelerometer + heart rate during sleep.

Citation: Schmidt, P. & Reiss, A. (2018). MMASH Dataset. PhysioNet. https://doi.org/10.24432/C57K5T

---

ML Model

Detail	Value
Architecture	LSTM(64) → Dropout(0.2) → LSTM(32) → Dropout(0.2) → Dense(1)
Input	30 timesteps × 6 features (movement magnitude, variance, avg HR, HRV, movement frequency, sleep duration)
Output	Sleep Score 0–100 → Good (≥70) / Poor (<70)
Format	TensorFlow Lite (.tflite) — optimized for edge inference
Scaler	MinMaxScaler fitted on training set, serialized as scaler.pkl

---

Project Structure

FogSleepMonitor/
├── dashboard/app.py              # Streamlit real-time dashboard
├── fog_node/fog_service.py       # Fog processing + TFLite inference
├── hardware/arduino_code/        # Arduino .ino firmware
├── models/
│   ├── sleep_model.tflite        # Optimized LSTM model (~64KB)
│   ├── sleep_lstm_model.h5       # Original Keras model
│   └── scaler.pkl                # MinMaxScaler
├── config.py                     # All settings
├── prepare_mmash_dataset.py      # Dataset preprocessor (MMASH)
├── train_model.py                # LSTM training pipeline
├── convert_to_tflite.py          # Keras → TFLite conversion
└── requirements.txt              # Python dependencies

---

Setup & Run

1. Install

git clone https://github.com/guhya-16/FogSleepMonitor.git
cd FogSleepMonitor
python -m venv .venv
.venv\Scripts\Activate.ps1       # Windows
pip install -r requirements.txt

2. Train (optional — pre-trained model included)

python prepare_mmash_dataset.py   # Downloads MMASH dataset
python train_model.py             # Trains LSTM
python convert_to_tflite.py       # Converts to TFLite

3. Hardware Wiring

MPU-6050	Arduino		Pulse Sensor	Arduino		Actuators	Arduino
VCC	5V		+	5V		LED (+)	Pin 13
GND	GND		–	GND		Buzzer (+)	Pin 8
SDA	A4		S	A0		Both (–)	GND
SCL	A5

Flash hardware/arduino_code/arduino_code.ino via Arduino IDE (baud: 115200).

4. Run

Terminal 1 — Fog Node:

$env:SLEEP_SERIAL_PORT="COM3"    # Set your port
python fog_node/fog_service.py   # Falls back to mock data if no Arduino

Terminal 2 — Dashboard:

streamlit run dashboard/app.py

Open http://localhost:8501 to view the live dashboard.

---

Why Fog Computing?

• Privacy — Health data stays on your machine, never uploaded to cloud
• Low Latency — No network round-trips, instant predictions
• Offline — Works without internet after model is trained
• Low Cost — Arduino (~₹500) + free Python stack

---

Tech Stack

Layer	Tools
Edge	Arduino UNO, MPU-6050, PPG Pulse Sensor
Fog	Python, TensorFlow Lite, NumPy, Pandas, Scikit-learn
Dashboard	Streamlit
Communication	PySerial (USB Serial)

---

Contributors

• @GuruMohith24
• @guhya-16

License

MIT