"C-IterSis: Consistent and Robust Unlimited Sampling"Guo, RuimingUnlimited Sensing Framework (USF) is a radically novel approach to sampling theory that focuses on the inverse problem of signal recovery from modulo-folded samples. The USF was developed to overcome fundamental bottlenecks inherent to the Shannon-Nyquist method, namely signal saturation or clipping and quantization noise. Bridging the gap between theory and practice, this work develops a novel reconstruction algorithm called Consistent Iterative Signal Sieving (C-IterSis). This method combines consistent sampling principle with an alternating minimization strategy to iteratively refine the signal recovery. The synergistic combination enables C-IterSis to effectively address issues like model mismatch, outliers, and system noise, all while achieving signal recovery at reduced sampling rates. We derive Cramér-Rao Bounds to benchmark the performance of the proposed algorithm. To demonstrate its efficacy in real-world conditions, C-IterSis was benchmarked against state-of-the-art methods and its utility is further corroborated through extensive hardware experiments (16 datasets) utilizing our proprietary modulo ADCs. This is joint work with Ayush Bhandari at Imperial College London. |
« back