Square-root Cubature Kalman Filter (SCKS)

This toolbox contains implementation of square-root Cubature Kalman Filter and square-root Rauch-Tang-Striebel smoother (SCKF-SCKS). These algorithms perform joint estimation of the states, input and parameters of stochastic continuous-discrete state-space models. The state equations must have a form of ordinary differential equations, where their discretization is performed through an efficient local-linearization scheme. Additionally, the parameter noise covariance is estimated dynamically via stochastic Robbins-Monro approximation method, and the measurement noise covariance is estimated online as well, using combination of varitional Bayesian (VB) approach with nonlinear filter/smoother. In particular, this method was designed to perform the nonlinear blind deconvolution of hemodynamic responses from fMRI data to estimate the underlying neuronal signal. Please contact Martin Havlicek (havlicekmartin@gmail.com) for any questions or suggestions.

Detailed description of this method can be found in our paper that has been accepted to NeuroImage: SCKSpaper


This software is distributed under the GNU General Public License (version 2 or later); please refer to the file License.txt , included with the software, for details.