Blind Separation of Instantaneous Mixtures of Sources Using Stochastic Calculus

Abstract

Stochastic calculus methods are used to estimate the Nonstationary components of a linear mixture. Each component in the mixture is assumed to follow a stochastic differential equation with some unknown parameters. The estimation of the stochastic process parameters and the estimation of the unknown amplitudes of the mixture matrix, through Girsanov theory, will generate a stochastic equation for each unknown process. Several examples are given, and a comparison to existing methods is provided.

Keywords: Ito Calculus, Blind Separation of Sources, Time-Series Analysis