ABSTRACT
Singular spectrum analysis in its various forms (multivariate, two-dimensional, tensor-based, hypercomplex, etc.) is largely dependent on the fundamental concept of subspace decomposition which has roots in the early mathematical development of signal, image, and information processing. In an intuitive and simple way, it introduces and involves the so-called embedding dimension in order to give freedom to the researcher to expand the number of subspaces to at least the number of actual components of interest. This is probably the main and most favourable concept which makes it different from its parent, principal component analysis.
