In this project, we are building an end-to-end clinical-grade IoT biometric monitor. We will use an ESP32 to capture raw electrocardiogram (ECG) and photoplethysmography (PPG) data, send it over Wi-Fi to a Python Flask backend, process 19 Heart Rate Variability (HRV) metrics, and use a Machine Learning ensemble to predict "Stress" or "No Stress" in real-time on a beautiful React.js dashboard.

The Vision & Social Impact:

• Chronic stress and workplace burnout represent a modern silent epidemic, directly contributing to cardiovascular diseases and mental health crises.
• Traditional assessment relies on subjective, retrospective questionnaires prone to human bias. Our system introduces a paradigm shift in behavioral health by creating an objective, continuous, and completely non-invasive stress tracking pipeline.
• By providing instantaneous physiological feedback, users can actively modify their behavior, practice real-time down-regulation techniques, and manage their well-being before a physical crisis occurs. 

• This project shifts healthcare from reactive treatment to proactive, data-driven self-awareness.

   

Commercialization & Productization Prospects:

This architecture is custom-built for high scalability and seamless transition into an entrepreneurship program.

• Corporate Wellness SaaS: The React dashboard can scale into a multi-user corporate portal where enterprises monitor anonymous, aggregated workforce stress levels to optimize workplace environments.
• Wearable Integration: The entire hardware footprints of the ESP32, AD8232, and MAX30102 can easily be miniaturized onto a single custom PCB, fitting into a lightweight chest strap or smartwatch prototype.

• B2B Licensing: The signal processing algorithm that converts raw values to 19 HRV metrics on a Flask server can be packaged as an API and licensed to existing health-tech applications.

   

Technology Practicality & Cost-Efficiency:

• Clinical ECG diagnostic equipment costs thousands of dollars and requires complex hospital setups.
• Our solution proves that high-performance medicine can be democratized. 

• Built entirely from off-the-shelf, low-cost commodity hardware totaling under $20, this system provides near-clinical observation accuracy at a fraction of the price, making continuous health tracking accessible to low-income populations and developing medical infrastructures.

   

Technical Architecture & Deep Design Detail:

The pipeline is engineered to extract clinical-grade biological markers on low-cost hardware through a three-tier architecture: 

1. The Edge Layer (MicroPython/ESP32): To prevent data loss during high-frequency sampling (~200Hz for ECG), we implemented an asynchronous dual-core threading architecture. Core 1 is dedicated entirely to precise sensor polling, while Core 0 handles Wi-Fi batching and network transmission.
2. The Computational Brain (Flask/Python): Raw bio-signals are prone to motion artifacts. The backend applies digital bandpass filtering to isolate R-peaks cleanly. Instead of calculating basic heart rate, the system evaluates a moving 60-second sliding window to extract 19 distinct time-domain and frequency-domain Heart Rate Variability (HRV) metrics.
3. The Ensemble Predictor: Rather than relying on a single classifier, we deployed a pre-trained Ensemble Machine Learning framework (trained on WESAD). This mitigates individual model biases, resulting in robust, reliable classifications ("Healthy", "Stressed") pushed to a React.js dashboard via WebSockets.

Components List:

Software Stack: 

• Micro-Python (Thonny)
• Python 3
• Flask
• Scikit-learn
• React.js

• Socket.io-client.

   

Circuit Connections:

Implementation:

1. Wire the Hardware: Connect the sensors to the ESP32 following the pinout table above. Ensure a common ground.
2. Flash the Edge Device: Upload the MicroPython firmware to the ESP32 using Thonny IDE.
3. Start the Brain: Run the Python Flask backend. It will begin listening on Port 5000 for the ESP32's TCP packets.
4. Launch the UI: Start the React.js development server to initialize the WebSocket connection with Flask.
5. Calibrate: Attach the ECG electrodes, place your finger on the MAX30102, and wait 60 seconds for the sliding window to fill and generate the first ML prediction.

Edge Programming:

• Reads the analog voltage from the AD8232.
• Rapidly polls the MAX30102 without blocking, checking for optical peaks to calculate BPM and SpO2.

• Batches this data and transmits it via HTTP POST exactly once every 5 seconds.

  

Machine Learning Inference:

• Sliding Window: It appends the live data into a 60-second rolling buffer.
• Signal Processing: It applies bandpass filters to the raw waveform, detects precise R-peaks, and computes 19 clinical Heart Rate Variability (HRV) metrics (like RMSSD, LF/HF ratio).

• Ensemble Machine Learning: It feeds these 19 metrics into a pre-trained Scikit-Learn Machine Learning pipeline (trained on the WESAD dataset) to instantly classify the data as "Healthy" or "Depressed."

  

Medical Dashboard:

• Sweeping line charts for ECG and optical pulse waveforms.
• Large metric cards for live BPM and Blood Oxygenation (SpO2).

• A dynamic UI status ring that glows Green for Healthy or Red for Stress based on the ML confidence score.

  

Testing and Demonstration:

• Attach the three AD8232 biomedical pads to your chest/arms.
• Place your index finger gently on the red glowing LED of the MAX30102.
• Start your Flask server and React app.

• Power up the ESP32.

   

Video:

Conclusion:

To conclude, by combining the raw hardware capabilities of the ESP32, the fast prototyping of MicroPython, and the predictive power of ensemble Machine Learning, we have created an open-source, edge-to-cloud biometric monitor. This architecture can be easily adapted for sports science, sleep studies, or personal mental health tracking.

Code: https://github.com/kathan-majithia/Smart_Physiological_Monitoring_System