Abstract

This paper presents the EEGraSP library, developed for processing and analyzing electroencephalogram (EEG) data using graph signal processing (GSP). The library is developed using the Python programming language scientific stack, including libraries such as NumPy, SciPy, Matplotlib, NetworkX, PyGSP2, and MNE. The development is open source, under the MIT license, and is done using the infrastructure available on GitHub. The respective documentation is also available, as are the mechanisms that facilitate the installation of the library through PyPI and conda-forge. The library offers three types of functionalities: the creation of graphs, either from the location of the electrodes or from time series associated with the electrodes (through graph learning algorithms); visualization of the graphs used to model the EEG data; and mechanisms to impute missing data in one or more electrodes. After describing the fundamentals of GSP and EEG acquisition and processing, the paper describes, through several examples, the different functionalities of EEGraSP. To the best of the authors’ knowledge, EEGraSP is the first library specifically designed for EEG analysis and processing using GSP. In the future, EEGraSP will include improvements in its development methodology, including unit testing within its continuous integration process, improvements in documentation and new examples, and the incorporation of new algorithms.