Bandlimited Signal Reconstruction From the Distribution of Unknown Sampling Locations

TL;DR

This paper investigates the challenges of reconstructing bandlimited signals from samples taken at unknown, statistically distributed locations, and proposes an estimation method when sample order is known, analyzing its accuracy and statistical properties.

Contribution

It demonstrates the impossibility of unique reconstruction without location order information and introduces an order-based estimation approach with proven asymptotic properties.

Findings

Reconstruction is impossible with unknown sample locations even with infinite samples.

Knowing the order of samples enables effective signal estimation.

The proposed estimator has quantifiable mean-squared error and asymptotic normality.

Abstract

We study the reconstruction of bandlimited fields from samples taken at unknown but statistically distributed sampling locations. The setup is motivated by distributed sampling where precise knowledge of sensor locations can be difficult. Periodic one-dimensional bandlimited fields are considered for sampling. Perfect samples of the field at independent and identically distributed locations are obtained. The statistical realization of sampling locations is not known. First, it is shown that a bandlimited field cannot be uniquely determined with samples taken at statistically distributed but unknown locations, even if the number of samples is infinite. Next, it is assumed that the order of sample locations is known. In this case, using insights from order-statistics, an estimate for the field with useful asymptotic properties is designed. Distortion (mean-squared error) and central-limit are established for this estimate.