A fixed-point algorithm for blind source separation with nonlinear autocorrelation (Q2378266)

Assume we observe a sensor signal \[ \mathbf{x(t)=As(t)}, \] where \(\mathbf{A}\) is unknown mixing matrix and \(\mathbf{s(t)}\) is a vector of unknown zero-mean and unit variance primary sources. The basic problem is to estimate both \(\mathbf{A}\) and \(\mathbf{s(t)}\).To that purpose the authors propose a fixed point algorithm based on nonlinear autocorrelations. The paper is complemented by a series of experimental results for both simulated and real time series. Especially the application to magnetoencephalographic recording might be of interest of readers.