ML-Based MMSE Source Separation

Simplified geometrical visualization of the optimality-gap between the estimation errors of the MMSE and the ML-based MMSE estimates.

This package contains three files: a (readme) detailed instruction file, the main function for the ML-based MMSE separation, and a script which demonstrates this function’s operation for an ICA problem with temporally-diverse stationary Gaussian sources.

To download the Matlab package, click here.

[1] Weiss, A. and Yeredor, A., “A Maximum Likelihood-Based Minimum Mean Square Error Separation and Estimation of Stationary Gaussian Sources from Noisy Mixtures”, IEEE Trans. on Signal Processing, vol. 67, no. 19, pp. 5032-5045, Oct. 2019. arXiv

[2] Weiss, A. and Yeredor, A., “Asymptotically Optimal Recovery of Gaussian Sources from Noisy Stationary Mixtures: The Least-Noisy Maximally-Separating Solution”, in Proc. of IEEE Int. Conference on Acoustics, Speech, and Signal Processing (ICASSP), pp. 5466–5470, May 2019.