Blind Sources Separation Based On Maximum Likelihood Approach

Blind sources separation (BSS) is a rising technology in recent years and has extensive and attractive application prospect in image processing、medical biological engineering and signal processing etc. Blind sources separation is a separation process that extract signal sources from a set of observed signals with unknown sources and mixture way. This paper uses maximum likelihood method to establish the problem’s model. We research the model’s properties and give the optimal solution conditions and the solution’s formula in two special situations by using matrix analysis and optimal method. Furthermore we prove the model corresponds to the following geometrical problem: given K ellipsoids in RK, find K vectors, w1∈ε1,···, wk∈εk, such that the volume of the parallelepiped defined by these vectors is maximum. Finally we design an algorithm for the model, some numerical experiments are performed showing the effectiveness of the algorithm.