Feature extraction from electroencephalographic records using EEGFrame framework

Analysis of electroencephalographic (EEG) signals usually includes visual inspection of the signal, feature extraction, and model generation. Computer-aided nonlinear feature extraction from EEG in particular has already led to improved descriptive and prognostic models of brain states and disorders. However, in this field, there is a lack of freely available powerful tools for scientific exploration of EEG that would help researchers to compare the results of their work with others. Especially, because of the great diversity of the proposed methods for EEG analysis, there exists a need for a joint framework for inspection, extraction and visualization performed on the EEG records. The aim of this paper is to introduce such a framework, called EEGFrame, with its implementation in Java. The framework currently supports the analysis of standard EDF records via signal inspection, feature extraction, and feature vectors storage for knowledge discovery. EEGFrame is the result of refactoring and extension of the HRVFrame framework for heart rate variability analysis, with added methods for EEG analysis. This paper describes the properties and capabilities of the framework and discusses its relevance with respect to similar work. The main advantage of EEGFrame is its support for numerous linear and nonlinear methods described in literature.