Real-Time ECG–PPG Feature Extraction (Arduino) — Stress Detection Path

This repository contains the Arduino-based real-time signal acquisition and feature extraction pipeline for a multimodal ECG + PPG stress detection prototype.

It represents the Real-Time Path in the overall workflow, where ECG and PPG signals are acquired from sensors, peaks are detected on-device, and features are computed every 10 seconds for downstream inference (external model / dashboard).

---

Purpose

The goal of this repo is to provide a working embedded pipeline to:

• acquire ECG (AD8232) and PPG (pulse sensor) signals in real time,
• perform robust peak detection (ECG R-peaks + PPG peaks),
• compute physiological features per 10-second window (aligned with the prototype requirements),
• optionally compute Pulse Arrival Time (PAT) as ECG→PPG delay,
• stream signals and detection markers for visual inspection via Serial Plotter.

This implementation was designed for a real-time experimental setup where 0-back (baseline / lower workload) and 2-back (higher workload) conditions were used as the two classes for stress / no-stress comparison (1-back was not used in this prototype setup).

---

Hardware Setup

• Arduino Mega
• ECG sensor: AD8232 (OUTPUT → A1, LO+/LO- optional)
• PPG sensor: Pulse sensor (signal → A0)
• Serial output used for:
  • live visualization (Serial Plotter),
  • streaming values for logging / dashboard integration.

---

What’s Included

1) Initial Signal Check (ECG + PPG)

A minimal sketch to confirm that both signals are being acquired and plotted.

• Reads ECG + PPG
• Prints both channels to Serial Plotter
• Useful for verifying wiring and sensor placement

2) Continuous Visualization + Peak Markers (Debug/Validation)

A visualization-focused sketch to validate that:

• PPG thresholding is working,
• QRS-enhancement for ECG helps detect R-peaks,
• beat markers appear correctly.

It streams:

• PPG
• PPG beat marker
• ECG
• QRS processed signal (scaled)
• ECG beat marker

This step was used to ensure signal quality and correct peak detection before final feature extraction.

3) Real-Time Feature Extraction (10-second windows)

A full real-time feature extraction sketch that:

• segments incoming data into 10-second windows,
• detects peaks with refractory control,
• computes features per window for both ECG and PPG,
• computes PAT (ECG R-peak → following PPG peak) when valid.

---

Extracted Features (per 10-second window)

PPG

• Mean inter-beat interval (IBI)
• IBI variance
• Mean heart rate (from PPG)
• Mean pulse amplitude (ADC range between beats)

ECG (HRV-style time-domain)

• Mean RR interval
• Mean heart rate (from ECG)
• SDNN
• RMSSD
• pNN50

PAT (ECG → PPG delay)

• PAT mean (ms)
• PAT variance
• Uses physiologically valid PAT range (e.g., 100–400 ms)

---

How to Run

1. Open the desired .ino sketch in Arduino IDE.
2. Select the correct board and port (Arduino Mega).
3. Upload the sketch.
4. Open Serial Plotter or Serial Monitor at 115200 baud.

For feature extraction:

• Send s to start collecting
• Send x to stop
• Features print automatically after each 10-second window

---

Notes

• Threshold values (PPG threshold, QRS threshold behavior) may need tuning depending on sensor placement and participant physiology.

• This repository covers only the Real-Time Path (embedded side) of the workflow attached below. Offline model training / dataset experiments / classification are maintained in a separate repository (https://github.com/Khadijakhanbme/ECG-PPG-Stress-Analysis)