/LSFT

Code for Lightweight Source Free Transfer for Privacy-Preserving Motor Imagery Classification

Primary LanguagePythonMIT LicenseMIT

Lightweight Source-Free Transfer for Privacy-Preserving Motor Imagery Classification

This repository contains codes of our paper https://ieeexplore.ieee.org/document/9840893

Datasets

The MI datasets can be downloaded and processed with MOABB in http://moabb.neurotechx.com/docs/datasets.html

Prerequisites:

  • python == 3.7.6

  • pyriemann == 0.2.6

  • mne == 0.20.7

  • numpy, scipy, sklearn

Running the code

Code files introduction:

main_demo_LSFT_voting.py -- demo file, LSFT with source models generated by voting.

dataloader.py -- function file, loading raw data and extract features.

data_align.py -- function file, perform centroid alignment.

feature_adaptation.py -- function file, perform feature adaptation on the virtual intermediate domain.