Effectiveness of Time-Warping in Signal Reconstruction from Nonuniformly Distributed Level-Crossing Samples

Johannes Königs, Carsten Bockelmann, Armin Dekorsy

Session: General

Keywords: Event-Based Sampling, Instantaneous Bandwidth, Time-warping

TL;DR: Analysis of how time-warping improves signal recovery from nonuniformly distributed level-crossing samples.

Abstract: The reconstruction of bandlimited signals from nonuniform samples remains a key challenge in modern signal processing, particularly for variable bandwidth (VBW) signals. Level-crossing (LC) sampling, a practical event-based sampling technique, adapts the sampling rate to the instantaneous bandwidth of a signal, but the inherent irregular sample spacing complicates signal reconstruction. In this paper, we analyze the effectiveness of time-warping in improving LC sample spacing for VBW signals and its impact on minimum energy (ME) reconstruction. By time-warping LC samples, we demonstrate a reduction in sample spacing variability, leading to a more stable reconstruction process. We evaluate the sample spacing distribution before and after time-warping and compare the reconstruction performance of ME reconstruction with and without time-warping. Numerical results show that time-warping significantly enhances reconstruction accuracy, reducing numerical instability in the ME method. This study highlights the potential of time-warping as an effective preprocessing step for LC sampled VBW signals.

Submission Number: 78