A low complexity estimation architecture based on noisy comparators

We consider a low-complexity architecture for scalar estimation using unreliable observations. A signal is observed using a number of binary comparisons for which the threshold levels can vary randomly. We analyze the statistics of this system and find a Cramér-Rao lower bound on the squared error performance of the estimator. By incorporating redundant observations and applying statistical estimation techniques, we form an estimate with error that is much smaller than the uncertainty in the threshold levels. We propose a two-stage architecture that achieves near-optimal mean square estimation error with low complexity. The performance of the architecture is evaluated using a simulated prototype.