Title :
Cramér-Rao Bound for Circular and Noncircular Complex Independent Component Analysis
Author :
Loesch, Benedikt ; Yang, Bin
Author_Institution :
Inst. of Signal Process. & Syst. Theor., Univ. Stuttgart, Stuttgart, Germany
Abstract :
Despite an increased interest in complex independent component analysis (ICA) during the last two decades, a closed form expression for the Cramér-Rao bound (CRB) for the demixing matrix is not known yet. In this paper, we fill this gap by deriving a closed-form expression for the CRB of the demixing matrix for instantaneous noncircular complex ICA. It contains the CRB for circular complex ICA and noncircular complex Gaussian ICA as two special cases. We also study the CRB numerically for the family of noncircular complex generalized Gaussian distributions and compare it to simulation results of two ICA estimators. Furthermore, we show how to extend the CRB to the case where the source signals are not temporally independent and identically distributed.
Keywords :
Gaussian distribution; independent component analysis; matrix algebra; Cramér-Rao bound; Gaussian ICA; Gaussian distributions; circular complex ICA; demixing matrix; independent component analysis; noncircular complex ICA; Algorithm design and analysis; Closed-form solutions; Covariance matrix; Independent component analysis; Indexes; Probability density function; Vectors; Cramér-Rao bound; generalized Gaussian distribution; independent component analysis; noncircular complex;
Journal_Title :
Signal Processing, IEEE Transactions on
DOI :
10.1109/TSP.2012.2226166