OECT Biosensor Signal Processing

Signal extraction, filtering, and noise reduction pipeline for a wearable cortisol level monitoring system based on PEDOT:PSS organic electrochemical transistors (OECTs). Developed as part of the OECT Biosensor Design project (LTspice, KiCad).

Signal model

Raw signal = baseline drift + cortisol response + thermal noise + flicker noise + 50Hz powerline

Processing pipeline

1. Bandpass filter (Butterworth, 0.05–200 Hz) — removes baseline wander
2. Notch filter (50 Hz, Q=30) — eliminates powerline interference
3. Baseline estimation — moving average over pre-stimulation window
4. Cortisol extraction — delta signal from baseline to steady state

Results

Metric	Value
SNR before filtering	~12 dB
SNR after filtering	~28 dB
Concentration error	< 5% across 10–200 nM range

How to run

pip install numpy scipy matplotlib
python signal_processing.py

Outputs

• signal_processing.png — raw vs filtered vs clean signal + FFT spectrum
• snr_vs_concentration.png — SNR across cortisol concentration range

Connection to quantum sensors

The same signal processing principles (SNR optimisation, noise modelling, filtering) apply directly to atomic sensors and magnetometry — key areas in quantum sensing and devices.

Skills demonstrated

Python, NumPy, SciPy signal processing, FFT, Butterworth/notch filters, SNR analysis, noise modelling