Optimal shift-invariant spaces from uniform measurements
Authors:
Rohan Joy,
Radha Ramakrishnan
Abstract:
Let $m$ be a positive integer and
$\mathcal{C}$ be a collection of closed subspaces in $L^2(\mathbb{R})$. Given the measurements $\mathcal{F}_Y=\left\lbrace \left\lbrace y_k^1 \right\rbrace_{k\in \mathbb{Z}},\ldots, \left\lbrace y_k^m \right\rbrace_{k\in \mathbb{Z}} \right\rbrace \subset \ell^2(\mathbb{Z})$ of unknown functions…
▽ More
Let $m$ be a positive integer and
$\mathcal{C}$ be a collection of closed subspaces in $L^2(\mathbb{R})$. Given the measurements $\mathcal{F}_Y=\left\lbrace \left\lbrace y_k^1 \right\rbrace_{k\in \mathbb{Z}},\ldots, \left\lbrace y_k^m \right\rbrace_{k\in \mathbb{Z}} \right\rbrace \subset \ell^2(\mathbb{Z})$ of unknown functions $\mathcal{F}=\left\{f_1, \ldots,f_m \right\} \subset L^2( \mathbb{R})$, in this paper we study the problem of finding an optimal space $S$ in $\mathcal{C}$ that is ``closest" to the measurements $\mathcal{F}_Y$ of $\mathcal{F}$. Since the class of finitely generated shift-invariant spaces (FSISs) is popularly used for modelling signals, we assume $\mathcal{C}$ consists of FSISs. We will be considering three cases. In the first case, $\mathcal{C}$ consists of FSISs without any assumption on extra invariance. In the second case, we assume $\mathcal{C}$ consists of extra invariant FSISs, and in the third case, we assume $\mathcal{C}$ has translation-invariant FSISs. In all three cases, we prove the existence of an optimal space.
△ Less
Submitted 2 April, 2025;
originally announced April 2025.
On the reconstruction of bandlimited signals from random samples quantized via noise-shaping
Authors:
Rohan Joy,
Felix Krahmer,
Alessandro Lupoli,
Radha Ramakrishnan
Abstract:
Noise-shaping quantization techniques are widely used for converting bandlimited signals from the analog to the digital domain. They work by "shaping" the quantization noise so that it falls close to the reconstruction operator's null space. We investigate the compatibility of two such schemes, specifically $ΣΔ$ quantization and distributed noise-shaping quantization, with random samples of bandli…
▽ More
Noise-shaping quantization techniques are widely used for converting bandlimited signals from the analog to the digital domain. They work by "shaping" the quantization noise so that it falls close to the reconstruction operator's null space. We investigate the compatibility of two such schemes, specifically $ΣΔ$ quantization and distributed noise-shaping quantization, with random samples of bandlimited functions. Let $f$ be a real-valued $π$-bandlimited function. Suppose $R>1$ is a real number and assume that $\{x_i\}_{i=1}^m$ is a sequence of i.i.d random variables uniformly distributed on $[-\tilde{R},\tilde{R}]$, where $\tilde{R}>R$ is appropriately chosen. We show that by using a noise-shaping quantizer to quantize the values of $f$ at $\{x_i\}_{i=1}^m$, a function $f^{\sharp}$ can be reconstructed from these quantized values such that $\|f-f^{\sharp}\|_{L^2[-R, R]}$ decays with high probability as $m$ and $\tilde{R}$ increase. We emphasize that the sample points $\{x_i\}_{i=1}^m$ are completely random, i.e., they have no predefined structure, which makes our findings the first of their kind.
△ Less
Submitted 1 July, 2023; v1 submitted 27 June, 2023;
originally announced June 2023.
