Optimal Signal and Detector Randomization under Average Power Constraints
Seminar by Sinan Gezici Department of Electrical and Electronics Engineering
Optimization of signal structures and detectors is studied for communications systems under power constraints in the presence of additive noise. First, it is observed that the optimal signal for each symbol can be characterized by a discrete random variable with at most two point masses. Then, optimization over all possible signal distributions and corresponding maximum a posteriori probability (MAP) decision rules are considered. It is shown that the optimization problem can be simplified into an optimization over a number of signal parameters instead of functions, which can be solved via global optimization techniques. Improvements that can be obtained via the joint design of the signaling and the detector are illustrated via examples. Finally, related problems such as optimal detector randomization and channel swithing are briefly discussed. (Joint work with Hui Lin)
