A weighted mixed statistics algorithm for blind source separation

Most blind source separation algorithms use either second order or higher order statistics in order to unmix the signals. In this paper we propose a novel weighted mixed statistics algorithm which performs significantly better than the single type statistics algorithms. Moreover, the algorithm is a generalisation of the single type statistics algorithm and requires thus less prior information. The weights are derived using the concept of estimating functions. Simulations are provided to show the enhanced performance of the weighted mixed statistics approach, even in mixtures were the signals contain no temporal information.