Second Order Blind Source Separation techniques (SO-BSS) and their relation to Stochastic Subspace Identification (SSI) algorithm

Abstract

Blind Source Separation (BSS) and Independent Component Analysis (ICA) techniques are emerging signal processing techniques that aim at identifying statistically independent sources from a linear mixture of such sources without requiring (or with little) a priori information about the input source signals. Recently, it has been shown that these techniques can also be utilized for Operational Modal Analysis (OMA), or Output-only Modal Analysis, where system characteristics are identified only on the basis of information available from the measured outputs. Second Order Blind Source Separation (SO-BSS) techniques are BSS algorithms that employ diagonalization of output correlation matrices in order to recover information about the original sources and the mixing matrix. Work presented in this paper aims at establishing a link between SO-BSS techniques (such as AMUSE and SOBI) and Stochastic Subspace Iteration (SSI) algorithm, which is a well known OMA algorithm. The paper presents the mathematical theory behind these algorithms and shows how these algorithms are related. In this manner, this work helps in enhancing the overall understanding of BSS techniques and their subsequent use for modal analysis purposes.KeywordsIndependent Component AnalysisModal ParameterIndependent Component AnalysisBlind Source SeparationBlind SignalThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.