|Stochastic Systems Group|
The talk introduces a new and efficient class of nonlinear detectors for vector symbols x chosen from a finite set. The detection is performed based on knowledge of a received vector r = Hx + w, where H is a matrix and w is a Gaussian noise vector. This model often appears in digital communication contexts, such as in the equalization of intersymbol interference (ISI) channels, the cancellation of multiple-access interference (MAI) in CDMA systems, and the decoding of multiple antenna systems.
The proposed ``iterated-decision'' detectors use optimized multipass algorithms to successively cancel interference from the received vector and generate symbol decisions whose reliability increases monotonically with each iteration. The thesis includes implementations when H is known at the receiver, when H and the statistics of w are unknown and must be learned at the receiver, and when channel coding is used. Strategies are also proposed to ensure that effectively optimal interference cancellation is achieved with arbitrary H.
This work is supervised by Prof. Gregory W. Wornell.
Problems with this site should be emailed to firstname.lastname@example.org