cvxeda 1.1.0

Python implementation of the cvxEDA algorithm

cvxEDA – Python implementation

This program implements the cvxEDA algorithm for the analysis of electrodermal activity (EDA) using methods of convex optimisation, described in:

A. Greco, G. Valenza, A. Lanata, E. P. Scilingo, and L. Citi
“cvxEDA: a Convex Optimization Approach to Electrodermal Activity Processing”
IEEE Transactions on Biomedical Engineering, 2015
DOI: 10.1109/TBME.2015.2474131

What the algorithm does

It is based on a model which describes EDA as the sum of three terms:

• phasic component – transient increases reflecting sudomotor bursts;
• tonic component – slowly varying baseline activity;
• additive white‑Gaussian‑noise term – captures model prediction errors, measurement noise and artifacts.

The model is physiologically inspired and fully explains EDA through a rigorous methodology based on Bayesian statistics, convex optimisation and sparsity.

Installation

# ── From PyPI (recommended)
pip install cvxeda                                # latest release
pip install "cvxeda==1.0.0"                       # or a specific version

# ── Directly from GitHub
pip install git+https://github.com/lciti/cvxEDA.git         # latest commit
pip install git+https://github.com/lciti/cvxEDA.git@v1.0.0  # or a specific version

# ── From a local clone
git clone https://github.com/youruser/cvxEDA.git  # clone
cd cvxEDA                                         # enter folder
git checkout v1.0.0                               # (optional) pick version
pip install .                                     # install

Quick example

import numpy as np
import cvxeda
import matplotlib.pyplot as plt

# Simulated EDA signal (replace with your own data)
tm = np.linspace(0, 40, 1000)                                           # 40 s at 25 Hz
y = np.convolve(np.sin(.7 * tm) < -.8, np.exp(-.5 * tm[:200]), 'same')  # simple fake signal
y += np.random.randn(len(y))                                            # add noise

yn = (y - y.mean()) / y.std()  # standardise
Fs = 25.0                      # sampling frequency (Hz)

# Run the algorithm
r, p, t, l, d, e, obj = cvxeda.cvxEDA(yn, 1.0 / Fs)

# Plot
tm = np.arange(1, len(y) + 1) / Fs
plt.figure()
plt.plot(tm, yn, label='normalised input')
plt.plot(tm, r,  label='phasic component')
plt.plot(tm, p,  label='sparse SMNA driver')
plt.plot(tm, t,  label='tonic component')
plt.xlabel('Time (s)')
plt.legend()
plt.show()

help(cvxeda.cvxEDA) prints the full function signature and argument description.

License & citation

The code is released under GPL‑3.0 (see LICENSE.txt).

If you use this program in support of published research, please cite the reference above. If you use this code in a software package, please explicitly inform the end users of this copyright notice and ask them to cite the reference above in their published research.

Project repository: https://github.com/lciti/cvxeda