Introduction

The project aims at providing Java-based portable software solution for the 2016 - 2019 Czech-Bavarian BCI project. It provides basic functionality to support:

• data acquisition (either off-line - from BrainVision files, or on-line - using Lab Streaming Layer API, or BrainVision RDA API),
• storing the data into a buffer
• segmentation (either into fixed-size segments for continuous data, or EEG marker-based epochs)
• pre-processing (such as channel selection, baseline removal, frequency filtering, and others)
• feature extraction (windowed means, downsampling, and others)
• classification (preferably using methods from deep learning category based on the Deeplearning4j library, such as stacked autoencoders)
• interpretation and evaluation of the results

Dependencies

Required libraries are handled by Maven. Sample off-line training and testing data based on the Guess the number experiment are a part of the project.

Directory structure

• src/main/java - contains source codes of the project structured into Java packages (note that preprocessing, feature extraction and classification methods are located in the cz.zcu.kiv.eeg.basil.processing.preprocessing, cz.zcu.kiv.eeg.basil.processing.featureextraction, cz.zcu.kiv.eeg.basil.processing.classification packages, respectively).

• src/main/resources - contains

  1. data/classifiers folder with stored classification configurations
  2. data/numbers folder with a large amount of data for classification training and testing from the 'Guess the number' experiment
  3. liblsl64.dll for LSL library communication

• src/main/test - contains tests for basic communication and a simple workflow test