DIWAH Intelligent Wearable Dashboard

Analytics dashboard for validating wearable sensor data against indirect calorimetry ground truth.

Compares accelerometer data from consumer devices (Bangle.js, EmotiBit) against research-grade Actigraph and metabolic measurements.

Prerequisites

• Python 3.9+ (Download)
• Docker (Download)

Quick Start

1. Clone and Setup

git clone <repository-url>
cd diwah-intelligent-wearable-dashboard

# Create virtual environment
python -m venv venv

# Activate (Windows PowerShell)
.\venv\Scripts\activate

# Activate (macOS/Linux)
source venv/bin/activate

# Install dependencies
pip install -r requirements.txt

2. Configure Environment

cp .env.example .env

Default .env works for local development:

INFLUX_URL=http://localhost:8086
INFLUX_TOKEN=<your-token>
INFLUX_ORG=diwah
INFLUX_BUCKET=wearables

3. Start InfluxDB

docker-compose up -d influxdb

4. Load Data

Option A: Import from backup

python scripts/import_influx_data.py <backup-folder>

Option B: Ingest from aligned CSV files

# Ingest pre-aligned data from output/aligned/
python scripts/ingest_aligned_data.py

# Verify
python scripts/verify_influx.py

5. Run Dashboard

python run_dashboard.py

Open: http://127.0.0.1:8050

---

Data Management

Import Data

python scripts/import_influx_data.py <backup-folder>

Verify Database

python scripts/verify_influx.py

---

Project Structure

├── src/
│   ├── analytics_dashboard.py   # Main Dash application
│   ├── backend/                 # Data loaders, parsers, correlation
│   ├── frontend/                # UI components, visualizations
│   └── config.py                # Configuration
├── scripts/
│   ├── ingest_aligned_data.py   # Ingest pre-aligned CSVs to InfluxDB
│   ├── export_influx_data.py    # Export to Line Protocol backup
│   ├── import_influx_data.py    # Import from backup
│   ├── verify_influx.py         # Database verification
│   └── reset_influxdb.py        # Wipe and recreate bucket
├── output/aligned/              # Pre-processed aligned data (5s epochs)
├── diwah-anonymized/            # Raw sensor data (not in git)
├── influxdb-backups/            # Data backups
└── docs/                        # Documentation

Data Pipeline

1. Raw Data (diwah-anonymized/) → Actigraph, Bangle, EmotiBit, Calorimetry files
2. Alignment (src/backend/alignment.py) → Synchronize timestamps, 5s resampling
3. Output (output/aligned/) → Pre-aligned CSVs per subject/device
4. Ingestion → Load to InfluxDB with subject/device/session tags
5. Dashboard → Query InfluxDB, visualize correlations

Documentation

• Participant Mapping — ID reference
• Alignment Pipeline — Synchronization logic
• Correlation Report — Validation results