PROJECT TITLE :

Double Coupled Canonical Polyadic Decomposition for Joint Blind Source Separation - 2018

ABSTRACT:

Joint blind supply separation (J-BSS) is an rising knowledge-driven technique for multi-set information-fusion. In this Project, J-BSS is addressed from a tensorial perspective. We show how, by using second-order multi-set statistics in J-BSS, a selected double coupled canonical polyadic decomposition (DC-CPD) downside can be formulated. We tend to propose an algebraic DC-CPD algorithm based mostly on a coupled rank-one detection mapping. This algorithm converts a probably underdetermined DC-CPD to a collection of overdetermined CPDs. The latter can be solved algebraically via a generalized eigenvalue decomposition based mostly scheme. Thus, this algorithm is deterministic and returns the precise solution in the noiseless case. Within the noisy case, it will be used to effectively initialize optimization based mostly DC-CPD algorithms. Additionally, we have a tendency to get the deterministic and generic uniqueness conditions for DC-CPD, that are shown to be more relaxed than their CPD counterpart. We tend to also introduce optimization based DC-CPD strategies, as well as alternating least squares, and structured data fusion based methods. Experiment results are given to illustrate the prevalence of DC-CPD over normal CPD primarily based BSS methods and many existing J-BSS strategies, regarding uniqueness and accuracy.