This repository is for EEG-bc project version control.
The project is to develop MATLAB codes for estimating Granger Causality in source space from EEG time series. Our group contains three members:
Parinthorn Manomaisaowapak and Anawat Nartkulpat and Jitkomut Songsiri Department of Electrical Engineering, Faculty of Engineering Chulalongkorn University, Bangkok, Thailand e-mail: [email protected], [email protected], [email protected]
The detail of mathematical formulation is described in our manuscript:
MS ID#: BIORXIV/2020/329276
MS TITLE: Granger Causality Inference in EEG Source Connectivity Analysis: A State-Space Approach
The folder contains data_generation: generate state-space models whose parameter correspond to have some sparse Granger causality pattern.
source_selection: estimation of state-space model with sparse prior on the rows of C (source output matrix).
gc_computation: calculate estimated Granger causality matrix.
pvo_subspace: original necessary files for subspace identification by Peter Van Overschee.
input_data: examples of EEG time series, head model, all required inputs to run experiment.
experiment: codes for each experiment explained in the paper.
Dependencies: Our codes rely on several sources in the following.
- Based codes for stochastic subspace identification by Peter Van OVerschee https://homes.esat.kuleuven.be/~smc/sysid/software/
- Model generation by S. Haufe et.al. This includes the calculation of head model and EEG signal generation. Our implementation extends the codes from Haufe. http://bbci.de/supplementary/EEGconnectivity/BBCB.html
- CVX: Our program require CVX installed in MATLAB (to solve the noise covariance estimation problem, which is a convex program) Available at http://cvxr.com/cvx/download/
- Some buit-in MATLAB commands from control or DSP toolboxes, e.g, solving RICCATI equation, generating pinknoise.