Skip to content

HappyFeat - Interactive framework for clinical BCI applications

docs status test: status PyPI version License: BSD-3

HappyFeat Logo

Testers and feedback wanted!

Reach out and give us your opinion, ideas for improvement, and to open issues for bugs or missing features!

HappyFeat is a software aiming to to simplify the use of BCI pipelines in clinical settings. More precisely, it is a software assistant for extracting and selecting classification features for BCI.

It gathers all necessary manipulations and analysis in a single convenient GUI, and automates experimental or analytic parameters. The resulting workflow allows for effortlessly selecting the best features, helping to achieve good BCI performance in time-constrained environments. Alternative features based on Functional Connectivity can be used and compared or combined with Power Spectral Density, allowing a network-oriented approach.

It consists of Qt-based GUIs and Python toolboxes, allowing to realize all steps for customizing and fine-tuning a BCI system: feature extraction & selection, classifier training.

HappyFeat also allows to interface with BCI softwares (OpenViBE and Timeflux) in order to facilitate the whole BCI workflow, from data acquisition to online classification.

The focus is put on ease of use, trial-and-error training of the classifier, and fast and efficient analysis of features of interest from BCI sessions.

HappyFeat Key Features

Warning

HappyFeat is still under development, and may undergo significant changes in the near future.

Requirements

  • Python 3.12
  • Python packages : shutils / PySide6 / numpy / MNE / matplotlib / scipy / spectrum / statsmodel / pandas / ruamel.yaml / plotly
  • OpenViBE Version 3.6.0: http://openvibe.inria.fr/downloads/
  • .. and/or Timeflux Version 0.16.0: https://doc.timeflux.io/en/stable/usage/getting_started.html#installation

Key Features

  • Easy to use GUI allowing to extract and visualize classification features, and select the most relevant ones for training a classifier.
  • Use Spectral Power or Coherence-based features for classification. HappyFeat allows to extract & visualize both types of features in parallel, and mix them at the training level.
  • Feature selection and classifier training can be done multiple times in a row, until satisfactory results are achieved.
  • A workspace management system keeps tracks of all extraction- and training-related manipulations, and enables a high degree of reproducibility.

Available pipelines:

  • With OpenViBE:
    • Power Spectrum (Burg method)
    • Functional Connectivity: Coherence-based Node Strength (Spectral Coherence, Imaginary part of Coherence)
    • Mixing Both PSD and Connectivity
    • Single-class classification
  • With Timeflux:
    • Power Spectrum (Welch method)
    • more coming soon...

Available classifiers:

  • LDA

Coming soon

  • More pipelines for Timeflux (Welch-based connectivity Node Strength)

  • More feature classes, more classification algos

  • Particle swarm algorithm for determining optimal AutoRegressive model order

  • Standalone Python software!

    • Run the offline analysis (extraction, visualization, selection, training) without the need of a BCI software