A new geometrical method of BSS on a lattice of the space of observations

Abstract

The techniques of Blind Separation of Sources (BSS) are used in many Signal Processing applications in which the data sampled by sensors are a mixture of signals from different sources, and the goal is to obtain an estimation of the sources from the mixtures. This work shows a new method for blind separation of sources, based on geometrical considerations concerning the observation space. This new method is applied to a mixture of several several sources and it obtains the estimated coefficients of the unknown mixture matrix A and separates the unknown sources. Following an introduction, we present a brief abstract of previous work on geometric ICA by our two groups, the principles of the new method and a description of the algorithm followed by some speed enhancements. Finally, we illustrate with simulations of several source distributions how the algorithm performs.