|Name||Separation of Digital Sources Mixtures by Convex Methods|
|Funding Reference||FCT - POSI/32948/CPS/2000|
The main objective is to design a spatial division multiple access (SDMA) receiver, which permits to blindly resolve a convolute mixture of digital sources. An existing convex geometrical framework for binary sources is to be extended in order to incorporate in the receiver the following features:
(a) Robustness to the channel order detection step. By (possibly) avoiding the estimation of the orders of the space-time channels activated by the sources, the receiver will outperform (in terms of robustness) most of the current approaches, whose performance is known to depend crucially on the accuracy of theses estimates – specially for band limited channels whose impulse response exhibits smooth roll offs making the channel length ill-defined. To meet this requirement, the linear equalizers have to be characterized directly in the observation space, which poses new theoretical challenges;
(b) Ability to cope with more generic digital modulation alphabets, e.g., PAM,QAM,PSK, etc;
(c) Increased robustness with respect to the noise power, which implies the formulation of new convex (or quasi-convex) problems in the “least-squares” sense.
Signal and Image Processing Group (SIPG)
Separation of Digital Sources Mixtures by Convex Methods