Nonparametric ICA for Nonstationary Instantaneous Mixtures

In this work, we use nonparametric sample Fisher information like matrix instead of the sample covariance matrix to find the mixing matrix in a nonstationary ICA model. This replacement may results in faster and robust separation.