Temporal principal component analysis - advances in dual auto-regressive modeling for blind Gaussian process identification

A recent paper (Cheung, 2001) has studied the blind identification of Gaussian source process through a general temporal independent component analysis (ICA) approach named dual autoregressive modelling. It is actually a temporal extension of the classical principal component analysis without considering the principal order of the components. In this paper, we further show the identifiable condition of the general temporal PCA (TPCA), and analyze the solution property of a specific TPCA algorithm presented in (Cheung 2001). Also, a new component ordering method is suggested, which includes the classical PCA ordering as a special case.