Abstract:In this article, we consider the random sampling in the image space
V$$ V $$ of an idempotent integral operator on mixed Lebesgue space
Lp,q()ℝn+1$$ {L}^{p,q}\left({\mathbb{R}}^{n+1}\right) $$. We assume some decay and regularity conditions on the integral kernel and show that the bounded functions in
V$$ V $$ can be approximated by an element in a finite‐dimensional subspace of
V$$ V $$ on
CR,S=[]−R2,R2n×[]−S2,S2$$ {C}_{R,S}={\left[-\frac{R}{2},\frac{R}{2}\right]}^… Show more
“…In the aforementioned papers, the construction of sampling stability strongly depends on two main tools, namely, covering number of signal space and large deviation inequality for a sequence of random variables. In fact, this is the case in most of the references on random sampling, such as previous research [10,11,18,19,[21][22][23][24][25][26]. The sampling stability can also be established based on the analysis of some localization operator and matrix Bernstein inequality; see previous studies [7,9,20] where the number of random samples to ensure sampling inequalities hold with high probability is at least of the order D ln D with D being the size of sampling domain.…”
Section: Introductionmentioning
confidence: 99%
“…random positions. In the last decade, random sampling has been investigated for bandlimited signals [7], signals in (quasi) shift-invariant spaces [9][10][11][18][19][20][21], signals with finite rate of innovation [22,23], and signals in reproducing kernel spaces [8,[24][25][26]. These papers establish sampling inequalities that hold with high probability when the number of sampling data is large enough.…”
Section: Introductionmentioning
confidence: 99%
“…These papers establish sampling inequalities that hold with high probability when the number of sampling data is large enough. Some of them also study the reconstruction from the exact sampling data by using sampling expansions or iterative algorithms [8, 23, 24] and from the convolution random samples by using sampling expansions [18, 19, 21]. In Section 4, we study the case that the sampling data are of the form (), and the sampling set consists of i.i.d.…”
Suppose that signals of interest reside in a reproducing kernel space defined on a metric measure space. We consider the scenario that the sampling positions are distributed on a bounded domain
of a metric measure space, and the sampling data are local averages of the original signals in a reproducing kernel space. For signals concentrated on
in that reproducing kernel space, we study the stability of this sampling procedure by establishing a weighted sampling inequality of bi‐Lipschitz type. This type of stability implies a weak version of conventional sampling inequality. We propose an iterative algorithm that reconstruct these concentrated signals from finite sampling data. The reconstruction error is characterized through the concentration ratio and the Hausdorff distance between the set of sampling positions and
. We also consider the random sampling scheme where the sampling positions are i.i.d. randomly drawn on
, and the sampling data are local averages of concentrated signals. We demonstrate that these concentrated signals can be approximated from the random sampling data with high probability when the sampling size is at least of the order
with
being the measure of
.
“…In the aforementioned papers, the construction of sampling stability strongly depends on two main tools, namely, covering number of signal space and large deviation inequality for a sequence of random variables. In fact, this is the case in most of the references on random sampling, such as previous research [10,11,18,19,[21][22][23][24][25][26]. The sampling stability can also be established based on the analysis of some localization operator and matrix Bernstein inequality; see previous studies [7,9,20] where the number of random samples to ensure sampling inequalities hold with high probability is at least of the order D ln D with D being the size of sampling domain.…”
Section: Introductionmentioning
confidence: 99%
“…random positions. In the last decade, random sampling has been investigated for bandlimited signals [7], signals in (quasi) shift-invariant spaces [9][10][11][18][19][20][21], signals with finite rate of innovation [22,23], and signals in reproducing kernel spaces [8,[24][25][26]. These papers establish sampling inequalities that hold with high probability when the number of sampling data is large enough.…”
Section: Introductionmentioning
confidence: 99%
“…These papers establish sampling inequalities that hold with high probability when the number of sampling data is large enough. Some of them also study the reconstruction from the exact sampling data by using sampling expansions or iterative algorithms [8, 23, 24] and from the convolution random samples by using sampling expansions [18, 19, 21]. In Section 4, we study the case that the sampling data are of the form (), and the sampling set consists of i.i.d.…”
Suppose that signals of interest reside in a reproducing kernel space defined on a metric measure space. We consider the scenario that the sampling positions are distributed on a bounded domain
of a metric measure space, and the sampling data are local averages of the original signals in a reproducing kernel space. For signals concentrated on
in that reproducing kernel space, we study the stability of this sampling procedure by establishing a weighted sampling inequality of bi‐Lipschitz type. This type of stability implies a weak version of conventional sampling inequality. We propose an iterative algorithm that reconstruct these concentrated signals from finite sampling data. The reconstruction error is characterized through the concentration ratio and the Hausdorff distance between the set of sampling positions and
. We also consider the random sampling scheme where the sampling positions are i.i.d. randomly drawn on
, and the sampling data are local averages of concentrated signals. We demonstrate that these concentrated signals can be approximated from the random sampling data with high probability when the sampling size is at least of the order
with
being the measure of
.
“…The invertibility or stability of the sampled representation is not investigated, however. In the context of random sampling, these properties are considered in the literature on relevant sampling, introduced by Bass and Gröchenig for bandlimited functions [28] and later generalized to various settings [29]- [31], including time-frequency representations [32]. Relevant sampling provides a probabilistic framework for stable sampling of functions that are localized in a domain of finite volume, e.g., bandlimited signals that have only negligible energy outside a finite interval.…”
The constant center frequency to bandwidth ratio (Q-factor) of wavelet transforms provides a very natural representation for audio data. However, invertible wavelet transforms have either required non-uniform decimation-leading to irregular data structures that are cumbersome to work with-or require excessively high oversampling with unacceptable computational overhead. Here, we present a novel decimation strategy for wavelet transforms that leads to stable representations with oversampling rates close to one and uniform decimation. Specifically, we show that finite implementations of the resulting representation are energy-preserving in the sense of frame theory. The obtained wavelet coefficients can be stored in a timefrequency matrix with a natural interpretation of columns as time frames and rows as frequency channels. This matrix structure immediately grants access to a large number of algorithms that are successfully used in time-frequency audio processing, but could not previously be used jointly with wavelet transforms. We demonstrate the application of our method in processing based on nonnegative matrix factorization, in onset detection, and in phaseless reconstruction.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.