Sampling Trajectories For The Short-Time Fourier Transform: Michael Speckbacher

Journal of Fourier Analysis and Applications (2022) 28:82
https://doi.org/10.1007/s00041-022-09977-9
Sampling Trajectories for the Short-Time Fourier Transform
Michael Speckbacher1
Received: 19 July 2021 / Revised: 22 September 2022 / Accepted: 5 October 2022

© The Author(s) 2022, corrected publication 2023
Abstract
We study the problem of stable reconstruction of the short-time Fourier transform
from samples taken from trajectories in R2 . We first investigate the interplay between
relative density of the trajectory and the reconstruction property. Later, we consider
spiraling curves, a special class of trajectories, and connect sampling and uniqueness
properties of these sets. Moreover, we show that for window functions given by a
linear combination of Hermite functions, it is indeed possible to stably reconstruct
from samples on some particular natural choices of spiraling curves.
Keywords Gabor analysis · Sampling theory · Mobile sampling · Polyanalytic

functions
Mathematics Subject Classification 42C15 · 42C30 · 42C40 · 65T60 · 94A20
1 Introduction
Reconstructing a function from incomplete data is an omnipresent task in fields like

signal processing or harmonic analysis. Classically, one aims to reconstruct a function
from discrete samples (see e.g. [34, 38]) and the stable version of this problem leads
to the notion of frames [12, 15]. For a variety of function spaces, the lower Beurling
density [10] (the definition being adapted to the particular geometry of the function
space in consideration) of the discrete sampling set needs to exceed a certain threshold
in order for the sampling process to form a frame, see, for example, [1, 24, 31, 33].
This threshold is commonly known as the Nyquist rate.
Communicated by Karlheinz Gröchenig.
B Michael Speckbacher
michael.speckbacher@univie.ac.at
1 Faculty of Mathematics, University of Vienna, Oskar-Morgenstern-Platz 1, 1090 Vienna, Austria
0123456789().: V,-vol
82 Page 2 of 23 Journal of Fourier Analysis and Applications (2022) 28:82
A more abstract form of the sampling problem is to characterize so called sampling

measures [6, 27, 28], i.e. measures μ on X that satisfy

p p
AF L p (X ) ≤ |F| p dμ ≤ BF L p (X ) ,
X
for some constants A, B > 0 and every F in a given subspace of L p (X ). A common

situation is that X = Rd and the measure takes the form μ = χΓ Hk , where Γ ⊂ X ,
χΓ is the characteristic function on Γ , and Hk , 0 ≤ k ≤ d, is the k-dimensional
Hausdorff measure. The case k = 0 corresponds to the classical discrete sampling
problem. For the second extreme case k = d, it turns out that a weaker notion of
density is often sufficient to stably reconstruct. For the Paley-Wiener space, this is
known as the Logvinenko-Sereda theorem [23, 25] and multiple extensions to other
spaces of analytic functions [18, 26, 28] and the range space of the short-time Fourier
transform [20] have since been established.
Recently, the study of the intermediate cases 0 < k < d on the Paley-Wiener space
P W2 (Ω) has drawn increasing attention and became known in the literature as the
mobile sampling problem [36, 37]. Mobile sampling has many applications whenever a
signal is measured by a moving sensor like in an MRI scan [8, 36]. Adapting Beurling’s
lower density for k = 1, the lower path density of a set Γ (measuring the limit average
length of the curve Γ in balls of increasing radii) was studied in [17]. In [22], it is
shown that a sufficiently large lower path density yields a stable sampling process. On
the other hand, there is no Nyquist rate that the lower path density needs to exceed for
stable sampling, see [17]. For particularly structured classes of trajectories Γ however,
Nyquist rates were established in terms of certain parameters that characterize the
’separation’ of Γ [21, 32].
In this paper, we study the equivalent of the mobile sampling problem for the
short-time Fourier transform

Vg f (z) = f (t)g(t − x)e−2πiξ t dt, z = (x, ξ ) ∈ R2 .
R
p
The modulation spaces Mϑ (R) are defined by

Mvp (R) := f ∈ S (R) : Vh 0 f · ϑ ∈ L p (R2 ) ,
where h 0 denotes the standard Gaussian, and ϑ is a submultiplicative weight, see,

p
e.g., [15]. We equip Mv (R) with the natural norm f M p = Vh 0 f · ϑ L p , and
ϑ
p
write M p (R) = Mϑ (R) if ϑ ≡ 1. This article is concerned with the study of Gabor
sampling trajectories for M p (R), that is, trajectory sets Γ ⊂ R2 for which

p p
A f M p ≤ |Vg f (z)| p dH1 (z) ≤ B f M p , f ∈ M p (R), (1)
Γ
holds. This is a particular instance of a sampling measure for the short-time Fourier
transform [6].
Journal of Fourier Analysis and Applications (2022) 28:82 Page 3 of 23 82
Fig. 1 Two examples of spiraling curves. Left The set of concentric circles Oη . Right The path
Sη (z 1 , . . . , z 4 ) generated by the points {z 1 , . . . , z 4 } = {(−a, a), (a, a), (a, −a), (−a, −a)}, a > 0
Just like in classical Gabor analysis, the stable sampling property relies heavily on
both the window g and the trajectory set Γ . On the one hand, one can quickly show that
for windows in M 1 (R), a necessary condition for Γ being a Gabor sampling trajectory
is that it is relatively dense [6] (see Sect. 3.1 for the definition of relative density). On
the other hand, applying a characterization of sampling measures on the Bargmann-
Fock space [28] we show that relative density is also sufficient for Gabor sampling
trajectories with Gaussian window (see Sect. 3.2). It is thus hopeless to search for a
Nyqist rate (both in terms of a path density as well as certain separation parameters)
that needs to be exceeded for general windows. However, in Proposition 4 we prove
that for certain windows a sufficient condition for Γ being a Gabor sampling trajectory
is that, for R ≤ Rg and every z ∈ R2 , H1 (Γ ∩ B R (z)) is bounded away from zero.
Later, we change perspective and study particular examples of trajectory sets. First,
we characterize sampling and uniqueness properties if Γ is a collection of parallel lines.
This will in turn be useful when studying the sampling property of a particular class
of trajectories: spiraling curves. Spiraling curves were introduced in [21] where their
sampling properties on the Paley-Wiener space were studied. In this article, we work
with a slightly more restrictive definition of spiraling curves that nevertheless includes
all the examples given in [21]. Loosely speaking, a spiraling curve is a trajectory set
that, when shifted in a particular direction, approaches a collection of equispaced
parallel lines or equispaced parallel edges in the limit. We spare the technical details
for now and refer to Definition 3 and Theorem 4 (Fig. 1).
Beurling’s theory of weak limits [10, 11, 34] establishes an equivalence between
sampling sets Γ for the Paley-Wiener space P W2 (Ω) and the uniqueness property
for all weak limits of translates of Γ on the Bernstein space P W∞ (Ω). This is a very
powerful result as the uniqueness property is often easier to verify. The price one has
to pay is that uniqueness has to be shown on a larger space and for all weak limits of
translates of the original sampling set. A similar result for sampling measures for the
modulation space M p (R) was shown by Ascensi [6]. In case of the measure χΓ dH1 ,
this result states that Γ is a Gabor sampling trajectory for M p (R) if and only if all
weak limits of translates of Γ are uniqueness sets for the weighted modulation space
p
M1/ϑs (R), for some s > 2 and ϑs (z) = (1 + |z|)s .
In most applications, Beurling’s theory is used to derive necessary conditions on

sampling sets. In [21], for example, the authors showed that certain weak limits of
translates of spiraling curves cannot be uniqueness sets for P W∞ (Ω) if the separation
exceeds a given threshold. Together with a result from [8], this yields a full character-
ization which sets of concentric circles and which Archimedes spirals are sampling
trajectories in terms of their separation parameter [21, Theorem A].
We take a different approach in this paper in that we fully characterize the set
of weak limits of translates of spiraling curves and then show (Theorem 4) that, for
windows taken from the linear span of the Hermite functions, the uniqueness property
is automatically satisfied for all weak limits of Γ except for Γ itself (and its finite
shifts). It then follows that Γ is a Gabor sampling trajectory for M p (R) if and only
p
if Γ is a uniqueness set for M1/ϑs (R), for some s > 2 (Theorem 5). Moreover, we
prove that certain spiraling curves are indeed such uniqueness sets. In particular, our
main result is the following.
Theorem 1 Let h n , n ∈ N0 , denote the Hermite functions (defined in (5)), g ∈
span{h n : n ∈ N0 }, 1 ≤ p < ∞, and η > 0. Moreover, let
(a) P be a star shaped polygon in R2 with kernel K such that the origin is contained
in the interior of K ,
(b) z 1 , . . . , z N ∈ R2 , N ≥ 3, be such that arg(z 1 ), . . . , arg(z N ), arg(z 1 ) is strictly
increasing or strictly decreasing modulo 2π .
The following sets are Gabor sampling trajectories for M p (R):
(i) the set of concentric circles Oη := {(x, y) ∈ R2 : x 2 + y 2 = η2 k 2 , k ∈ N},
(ii) the collection of star shaped polygons Pη := {(x, y) ∈ R2 : (x, y) ∈ ηk P, k ∈
N},
(iii) Sη (z 1 , ..., z N ), the path generated by the sequence of vectors
{ηz 1 , ηz 2 , ..., ηz N , 2ηz 1 , 2ηz 2 , ..., 2ηz N , 3ηz 1 , 3ηz 2 , ...}.
This paper is organized as follows. After a section dedicated to preliminaries, we study

some basic examples of windows and trajectory sets and give necessary and sufficient
conditions for Gabor sampling trajectories in that setting in Sect. 3. Then in Sect. 4 we
introduce spiraling curves and study their weak limits of translates which then allows
us to prove our main result.
2 Preliminaries
2.1 Notation
Throughout this paper we adopt the following conventions. We write d for vectors in
S1 and (θ ) := (cos(2π θ ), sin(2π θ )), θ ∈ T ∼ = [0, 1) as well as ei , i = 1, 2, for
the standard basis vectors in R2 . For d ∈ S1 we write d⊥ for the vector obtained by
rotating d clockwise by π/2. The space of continuous, compactly supported functions
is denoted by Cc (Rd ), the ball of radius R and center z by B R (z), and we write A B
if there exist C > 0 such that A ≤ C B.
2.2 Hausdorff Measures and Trajectory Sets
Let E ⊂ R2 . The 1-dimensional Hausdorff measure of E is given by

⎧ ⎫
⎨ ⎬
H1 (E) = lim inf 2 rj : E ⊂ B(x j , r j ) and r j ≤ δ .
δ0 ⎩ ⎭
j j
When restricted to a line, H1 equals the 1-dimensional Lebesgue measure.

Let ϕ : [0, 1) → [0, ∞) be continuous at 0. A (locally finite Borel) measure μ on
R2 is called ϕ-regular if for every z ∈ R2 and every R ∈ (0, 1) one has
μ(B R (z)) ≤ π ϕ(R)R.
A set E ⊂ R2 is called ϕ-regular if H1 | E is ϕ-regular.

A trajectory Γ ⊂ Rd is the image of a curve γ : R → Rd , i.e. Γ = γ (R), such that
each restriction of γ to a finite interval is rectifiable. A trajectory set Γ is a countable
collection of trajectories.
2.3 The Short-Time Fourier Transform and Modulation Spaces
Let z = (x, ξ ) ∈ R2 . A time-frequency shift of a function g is given by
π(z)g(t) = Mξ Tx g(t) = e2πiξ t g(t − x).
Given a window g ∈ L 2 (R), the short-time Fourier transform of f ∈ L 2 (R) is defined

as

Vg f (z) = f , π(z)g = f (t)g(t − x)e−2πiξ t dt.
R
The short-time Fourier transform satisfies the following orthogonality relation

Vg1 f 1 (z)Vg2 f 2 (z) dz = f 1 , f 2 g1 , g2 ,
R2
which implies that, if g2 = 1, Vg : L 2 (R) → L 2 (R2 ) is an isometry.

Let ϑ : R2 → R+ be a submultiplicative weight function, i.e. ϑ(z + w) ≤
p
ϑ(z)ϑ(w). The (weighted) modulation spaces Mϑ (R), 1 ≤ p ≤ ∞, can be defined
by

Mϑ (R) := f ∈ S (R) : Vh 0 f · ϑ ∈ L p (R) ,
p
where h 0 denotes the normalized Gaussian and S (R) the space of tempered distri-
butions. Modulation spaces are Banach spaces when equipped with the natural norm
p
f M p = Vh 0 f · ϑ L p . If ϑ ≡ 1, we write M p (R) = Mϑ (R). See, for example, [9,
ϑ
15] for a thorough introduction to the topic. Throughout this paper we only consider
polynomial weight functions, i.e.,
ϑs (z) = (1 + |z|)s , z ∈ R2 , s ≥ 0.
We note here that Mϑ1s (R) is closed under pointwise multiplication with functions
from the weighted Fourier algebra Aϑs (R)

Aϑs (R) := f ∈ C0 (R) : |
f (ξ )|ϑs (ξ ) dξ < ∞ .
R
In particular, for f ∈ Aϑs (R) and g ∈ Mϑ1s (R) we have
f g M 1 ≤ f Aϑs g M 1 , (2)

ϑs ϑs
see, for example, the arguments in [19, Proposition 4.13] which can easily be adapted
to the weighted case.
Let g ∈ M 1 (R). We call a trajectory set Γ ⊂ R2 a Gabor sampling trajectory for
M (R) if there exist constants A, B > 0 such that
p

p p
A f M p ≤ |Vg f (z)| p dH1 (z) ≤ B f M p , f ∈ M p (R). (3)
Γ
If only the upper bound is satisfied, then we call Γ a Gabor Bessel trajectory. If p = 2,
then M p (R) = L 2 (R), and (3) is equivalent to {π(z)g}z∈Γ forming a continuous
frame, see [5, 30].
The trajectory set Γ is called a uniqueness set for M p (R), if Vg f |Γ = 0, f ∈
M (R), implies f = 0.
p
2.4 Hermite Windows and Polyanalytic Functions
A function F : C → C is called polyanalytic of order n if it satisfies the higher order

¯ n+1 F = 0. In that case, F can be written as
Cauchy-Riemann equation (∂)

n
F(z) = F(z, z) = Fk (z)z k , z ∈ C, (4)
k=0
where F0 , . . . , Fn : C → C are holomorphic functions.

The Hermite functions are given by
n
21/4 −1 d n −2π t 2
eπ t
2
h n (t) = √ √ (e ), n ∈ N0 . (5)
n! 2 π dt n
p
For f ∈ Mϑs (R), 1 ≤ p < ∞, the function F given by
F(z) = Vh n f (z)eπ(z
2 −z 2 )/4
eπ |z|
2 /2
, (6)
is polyanalytic of order n [2], where we identify z = (x, ξ ) ∈ R2 with z = x +iξ ∈ C.

n π(z 2 −z 2 )/4 eπ |z|2 /2 is polyanalytic of
In particular, if g = k=0 αk h k , then Vg f (z)e
order n.
A polyanalytic function F of order n is called reduced if it can be written as

n
F(z) = Fk (z)|z|2k , Fk holomorphic.
k=0
Reduced polyanalytic functions satisfy the following Cauchy-type formula [7, Sect.
1.3, (11)]. For a similar Cauchy-type formula for true polyanalytic functions, we refer
to [4].
Lemma 1 (Balk) Let F be a reduced polyanalytic function of order n in B R (0), 0 <
R0 < R1 < . . . < Rn < R, and let Γk := {z : |z| = Rk }. For every z ∈ B R0 (0)

1
n
F(t)
F(z) = Pk (|z|2 ) dt, (7)
2πi
k=0 Γk − z
t
R 2j −t
where Pk (t) := j=k R 2 −R 2 .
j k
2.5 The Metaplectic Rotation

cos(2πθ) − sin(2πθ)
Let us denote the rotation matrices in R2 by R(θ ) = sin(2πθ) cos(2πθ) , θ ∈ T. The
metaplectic rotation of f ∈ L 2 (R) is given in terms of the Hermite basis {h n }n∈N0

U (θ ) f := e−2πinθ f , h n h n .
n≥0
Clearly, U (θ ) is a unitary operator on L 2 (R) with U (θ )∗ = U (−θ ), and U (θ )h n =

e−2πinθ h n . For f , g ∈ L 2 (R), the standard rotation of the argument of the short-time
Fourier transform and the metaplectic rotation are connected via the formula

Vg f (R(θ )z) = eπi(xω−x ω ) VU (θ)g U (θ ) f (z), z = (x, ω), and (x ω ) = R(θ )z,
(8)

or equivalently via π(R(θ )z) = e−πi(xω−x ω ) U (−θ )π(z)U (θ ). This is a special case
of the symplectic covariance of the Schrödinger representation, see [13, Chaps. 1 and
2], and [15, Chap. 9].
As ϑs is radially symmetric, (8) implies that f M 1 = U (θ ) f M 1 , which allows

ϑs ϑs
∞ (R) = M 1 (R)∗ via
to extend the operator U (θ ) to M1/ϑs ϑs
∞
U (θ ) f , g M ∞ = f , U (−θ )g M ∞ , f ∈ M1/ϑ (R), g ∈ Mϑ1s (R).
1/ϑs ×Mϑs 1/ϑs ×Mϑs
1 1
s
∞ (R), and g ∈ M 1 (R).

Consequently, (8) remains valid for f ∈ M1/ϑs ϑs
3 General Trajectory Sets
In this section, we present some basic necessary and sufficient conditions for Gabor
sampling trajectories. Moreover, we study the cases of (i) sampling on general trajec-
tory sets for specific windows, in particular, the Gaussian window (Corollary 1) and a
certain class of window functions (Proposition 4), and (ii) sampling on parallel lines
for general windows (Proposition 2).
3.1 Relative Density
We call a trajectory set Γ ⊂ R2 (m, R)-dense if
inf H1 (Γ ∩ B R (z)) ≥ m > 0,

z∈R2
and relatively dense if there exist constants m, R > 0 such that Γ is (m, R)-dense. It
turns out that relative density is a necessary condition for Γ being a Gabor sampling
trajectory for M p (R), see [6, Theorems 8 and 10]:
Proposition 1 (Ascensi) Let g ∈ M 1 (R) and 1 ≤ p < ∞. If Γ ⊂ R2 is a Gabor

Bessel trajectory for M p (R), then there exist M, R > 0 such that
sup H1 (Γ ∩ B R (z)) ≤ M. (9)

z∈R2
If Γ is a Gabor sampling trajecory for M p (R), then Γ is relatively dense.
Note that the proof of (9) can easily be adapted to the case p = 2 and general g ∈ L 2 (R)
and explicit upper bounds of the Bessel constant can be derived, for example, from
[3, 4].
The question whether relative density is also necessary for general windows and
Gabor sampling trajectories for L 2 (R) remains open. Drawing comparison to the
discrete [31] and planar cases [20] however suggests that relative density should indeed
be a necessary requirement.
3.2 Gaussian Window
In [28], Ortega-Cerdà fully characterized the sampling measures for the Bargmann-
Fock space of entire functions which corresponds to the short-time Fourier transform
with Gaussian window h 0 . The result goes as follows.
Theorem 2 (Ortega-Cerdà) Let 1 ≤ p < ∞. The measure μ is a sampling measure
for Vh 0 (M p (R)) if and only if there exist constants R, δ, M > 0 and N ∈ N such that
n(R, N , δ, z)
(i) sup μ(B R (z)) ≤ M, (ii) inf > 1,
z∈R2 z∈R2 R2
where n(R, N , δ, z) is calculated by the following rule: take S R (z) := z +

[−R/2, R/2)2 and cover it with N 2 smaller squares of sidelength R/N . Then
n(R, N , δ, z) denotes the number smaller squares s ⊂ S R (z) that satisfy μ(s) ≥ δ.
When specifying the measure μ to be the 1-dimensional Hausdorff measure on Γ , we
subsequently show that, with a minor extra condition on the trajectory set Γ , condition
(ii) is equivalent to Γ being relatively dense.
Corollary 1 Let g = h 0 be the standard Gaussian window, 1 ≤ p < ∞, ϕ : [0, 1] →
[0, ∞) be continuous at 0, and Γ ⊂ R2 be a ϕ-regular trajectory set. Then Γ is a
Gabor sampling trajectory for M p (R) if and only if there exist R, m, M > 0 such
that
m ≤ H1 (Γ ∩ B R (z)) ≤ M, for every z ∈ R2 . (10)
Proof The second inequality of (10) is condition (i) in Theorem 2. Hence, we have to
show the equivalence of the left hand side inequality and (ii). If (ii) holds, then
H1 (Γ ∩ B R (z)) ≥ H1 (Γ ∩ S R (z)) ≥ δn(R, N , δ, z) > δ R 2 .
Now, let Γ be relatively dense. Note that we may assume that ϕ(0) > 0 (for if Γ is
ϕ-regular,√then it is also (ϕ + ε)-regular
√ √ ε > 0). Let us choose N large enough
for any
such that 2R/N < 1, and ϕ( 2R/N ) ≤ 2ϕ(0). Then, using the lower bound in
(10) and the ϕ-regularity of Γ , we get
m ≤ H1 (Γ ∩ B R (z)) ≤ H1 (Γ ∩ S2R (z))

≤ n(2R, N , δ, z) sup H1 (Γ ∩ S 2R (w)) + (N 2 − n(2R, N , δ, z))δ
N
w∈S2R (z)
R
≤ n(2R, N , δ, z)2π ϕ(0) + (N 2 − n(2R, N , δ, z))δ.
N
Regrouping the terms yields and choosing δ = δ N = 1

N2
min{m/2, π Rϕ(0)N } shows
that
m − N 2δ m
n(2R, N , δ, z)) ≥ N ≥N ,
2π Rϕ(0) − δ N 4π Rϕ(0)
which can be arbitrarily large as N → ∞. In particular, there exists N ∈ N such that

(ii) is satisfied.

Remark 1 In [29] it is shown that ϕ-regularity with ϕ(0) = 1 can be considered as a
quantitative strengthening of rectifiability. On the other hand, it is possible to construct
fractal sets that are ϕ-regular if ϕ(0) > 1, which shows that ϕ-regularity for some ϕ
is a rather mild assumption.
3.3 The Case of Parallel Lines
We now study sampling and uniqueness properties of sets of parallel lines which
follow from simple arguments. The results however will be useful later when we
study spiraling curves.
Proposition 2 Let g ∈ L 2 (R), Λ ⊂ R be countable, and d = R(θ )e2 . The collection
of parallel lines L d,Λ , where

L d,Λ := t d + λd⊥ : t ∈ R, λ ∈ Λ ,
is a Gabor sampling trajectory for L 2 (R) with sampling bounds A, B > 0 if and only
if

A≤ |U (θ )g(t − λ)|2 ≤ B, for a.e. t ∈ R. (11)
λ∈Λ
Proof First, by (8) we may rotate the problem and assume d = e2 and d⊥ = e1 .
Writing the short-time Fourier transform as Vg f (x, ξ ) = F f Tx g (ξ ) and using
Parseval’s identity yields

|Vg f (z)|2 dH1 (z) = |Vg f (λ, ξ )|2 dξ = | f (t)g(t − λ)|2 dt
L e2 ,Λ λ∈Λ R λ∈Λ R

= | f (t)| 2
|g(t − λ)| 2
dt,
R λ∈Λ
where changing the order of integration and summation is allowed by either (11) or
the existence of the upper sampling bound.

∞ (R) can be
The uniqueness property of parallel lines on the distribution space M1/ϑs
characterized in a similar fashion.
Proposition 3 Let g ∈ Mϑ1s (R) ∩ Aϑs (R), s ≥ 0, Λ ⊂ R be countable, and d =
∞ (R) if and
R(θ )e2 . The collection of parallel lines L d,Λ is a uniqueness set for M1/ϑs
only if
supp(Tλ U (θ )g) = R,

λ∈Λ
where
supp(g) := {t ∈ R : g(t) = 0} denotes the effective support of g.
Remark 2 In [6, Lemma 27], a similar result was shown for planar uniqueness sets.
Since some technical details are left out there, we decided to include the proof here.
Proof For f ∈ M1/ϑ ∞ (R), g ∈ M 1 (R) ∩ A (R) and h ∈ M 1 (R), we define the
s ϑs ϑs ϑs
product f g via f g, h M ∞ ×M 1 = f , gh M ∞ ×M 1 . By (2), it then follows that
1/ϑs ϑs 1/ϑs ϑs
∞ (R) as
f g ∈ M1/ϑs

f g, h ∞ ×M 1 = f , gh ∞
1/ϑs ×Mϑs
M1/ϑ M 1
s ϑs
≤ f M1/ϑ
∞ gh 1 ≤ f M ∞ g A h 1 .
M ϑs M
s ϑs
1/ϑ s ϑs
∞ ⊂
Moreover, since Aϑs (R) is invariant under translations it follows that f Tx g ∈ M1/ϑ
s
S (R) and consequently, F f Tx g ∈ S (R) is a well defined tempered distribution.
∞ (R), and since the short-time Fourier trans-
Since Mϑ1s (R) is weak-∗ dense in M1/ϑ s
form of a distribution in M1/ϑ∞ (R) is continuous in R2 , it follows that V f (x, ξ ) =
g
s
F f Tx g (ξ ).
As before we may rotate the problem and assume d = e2 . Now, Vg f | L e2 ,Λ = 0 if
and only if all the distributions f Tλ g, λ ∈ Λ, are zero. This in turn is equivalent to
the support of f and the effective support of Tλ g being disjoint for every λ ∈ Λ.
3.4 Connection to Discrete Sampling
Proposition 2 shows that Γ being relatively dense is not sufficient for Γ to be a Gabor
sampling trajectory for L 2 (R) = M 2 (R). A natural question is therefore whether for
every g ∈ L 2 (R)\{0} there exists R ∗ = R ∗ (g) > 0 such that every (γ , R)-dense
trajectory set Γ is a Gabor sampling trajectory if R ≤ R ∗ . We follow the approach of
[20] to show that such R ∗ does in fact exist for a certain class of window functions.
Let us write Q R (z) := Rz+[−R/2,
R/2)2 , Xg(t) = tg(t), and recall the definition

of the Sobolev space H 1 (R) := f ∈ L 2 (R) : R (1 + |ξ |2 )| f (ξ )|2 dξ < ∞ .
Proposition 4 Let g, tg ∈ H 1 (R), R > 0 be chosen such that

2R 2R
Δ := g 2 + Xg2 + Xg 2 < g2 , (12)
π π
and Γ ⊂ R2 be a trajectory set. If there exist m, M > 0 such that
0 < m ≤ H1 (Γ ∩ Q R (z)) ≤ M < ∞, for every z ∈ R2 , (13)
then for every f ∈ L 2 (R)

2 2
m g2 − Δ f 2 ≤ |Vg f (z)|2 dH1 (z) ≤ M g2 + Δ f 2 .
Γ
Proof It is shown in [35] that, for the particular choice of g and R, arbitrary points
z n ∈ Q R (n), n ∈ Z2 , generate a discrete frame {π(z n )g}n∈Z2 for L 2 (R) with uniform
2 2
frame bounds A = g2 − Δ and B = g2 + Δ .
For every n ∈ Z2 there exists z n ∈ Γ ∩ Q R (n) such that

1
|Vg f (z n )|2 ≥ |Vg f (z)|2 dH1 (z).
H1 (Γ ∩ Q R (n)) Γ ∩Q R (n)
Then, as every choice of points z n ∈ Q R (n) generates a Gabor frame with uniform
upper bound B, we have

|Vg f (z)|2 dH1 (z) = |Vg f (z)|2 dH1 (z)
Γ Γ ∩Q R (n)
n∈Z2

≤ H1 (Γ ∩ Q R (n))|Vg f (z n )|2 ≤ M B f 2 .
n∈Z2
The lower sampling bound follows with a similar argument.

Remark 3 This construction works for general measures and gives a characterization
of sampling measures for this class of window functions.
As Q R (z) ⊂ B√2R (z) ⊂ Q 2R (z), it follows that (13) implies m ≤ H1 (Γ ∩
B√2R (z)) ≤ 4M, for every z ∈ R2 , that is, Γ is relatively dense in the sense of
Sect. 3.1. The quantitative estimate of the frame bounds however depends on the
relation (13).
4 Spiraling Curves
4.1 Weak Limits
There are multiple ways of defining weak limits of trajectory sets. The definition in
[21], for example, adapts the original notion by Beurling [11] given in terms of a
geometric condition. For our purposes, it will be more convenient to work with a
stronger notion that was introduced to define weak limits of measures, see [6].
Definition 1 Let Γ , Γ , Γn be trajectory sets in R2 . We say that {Γn }n∈N converges

weakly to Γ if

φ(z) dH1 (z) → φ(z) dH1 (z),
Γn Γ
w
for every nonnegative function φ ∈ Cc (R2 ). In that case we write Γn → Γ .
We say that Γ is a weak limit of translates of Γ if there exists a sequence {z n }n∈N
w
such that z n + Γ → Γ and define WΓ as the set of all weak limits of translates of Γ .
The following characterization of Gabor sampling trajectories is an immediate conse-

quence of a characterization of samping measures for the short-time Fourier transform
given by Ascensi [6, Theorem 14]. One can think of this result as a time-frequency
analog of the classical result by Beurling [11, Theorem 3, p. 345] where the Gabor
sampling property on M p (R) is connected to the uniqueness property of all weak limits
p
of translates on the larger space M1/ϑs (R). Note that Ascensi’s result requires a spe-
cial class of windows which we define in a simplified version taylored to polynomial
weights.
Definition 2 We say that 1/ϑs controls Vg g if there exists a decreasing function d :
R+ → R+ with d(r ) → 0, as r → ∞, such that |Vg g(z)| ≤ d(|z|)/ϑs (z). The class
M(R) is then given by

M(R) := g ∈ L 2 (R) : ∃s > 2 s.t. 1/ϑs controls Vg g .
Note that M(R) contains, for example, functions in the Schwartz class as well as
window functions considered in the theory of intrinsically localized frames [14].
Theorem 3 (Ascensi) Let g ∈ M(R), s > 2 be such that 1/ϑs controls Vg g, and
1 ≤ p < ∞. Then Γ is a Gabor sampling trajectory for M p (R) if and only if every
Γ ∈ WΓ is a set of uniqueness for M1/ϑs (R).
p
4.2 Spiraling Curves and Their Weak Limits
The notion of a spiraling curve was introduced in [21]. This class of trajectory sets
includes a wide range of natural examples such as the concentric circles or the
Archimedes spiral. In this paper, we use a slightly more restrictive notion of spi-
raling curves that still includes the main examples from [21] while allowing for a full
characterization of the set of weak limits of translates.
Definition 3 (Spiraling curve) Let IΓ ⊂ T be a finite set. A trajectory set Γ is called

spiraling if the following conditions are satisfied:
If β ∈ T\IΓ , then
(A.i) (Escape Cone) for α := min{dist(β, IΓ ), 1/8} the intersection of Γ with the
cone
Cα,β := {(r cos(2π θ ), r sin(2π θ )) : r ≥ 0, β − α ≤ θ ≤ β + α}
can be parametrized in polar coordinates as
γβ (θ ) = (rβ (θ ) cos(2π θ ), rβ (θ ) sin(2π θ ))

with θ ∈ k∈N [k + β − α, k + β + α] and rβ a nonnegative C 2 -function on
each interval whose C 2 -norm is globally bounded.
(A.ii) (Asymptotic radial monotonicity) there exists K ∈ N such that for any θ ∈
[−α, α] fixed, the sequence rβ (k + β + θ ) is strictly increasing for k ≥ K .
(A.iii) (Asymptotic flatness) the curvature of γβ , denoted by κβ , tends to 0 as its input

goes to infinity. To be more precise, we assume
sup κβ (k + β + θ ) → 0, k → ∞. (14)
θ∈(−α,α)
(A.iv) (Asymptotic equispacing) there exist continuous functions ηβ : [−α, α] →

R+ , ρβ : [−α, α] → R such that

sup rβ (k + β + θ ) − ηβ (θ )k − ρβ (θ ) → 0, k → ∞.
θ∈[−α,α]
(A.v) (Asymptotic velocity) there exists a continuous function dβ : [−α, α] → S1

such that dβ (θ ) is non-collinear with (θ ) and

γ (k + β + θ )
β
sup − dβ (θ ) → 0, k → ∞. (15)
θ∈[−α,α] γβ (k + β + θ )
If β ∈ IΓ , then (A.i) − (A.v) hold with the following modifications

(B.i) for α := min{dist(β, IΓ \{β}), 1/8}, the parametrization γβ |[k+β−α,k+β+α] is
continuous and the restrictions of γβ to (k + β, k + β + α]
and [k + β − α, k + β) are C 2 -functions with bounded C 2 -norms.
(B.iii) The supremum in (14) is replaced by an supremum over (−α, α)\{0}.
(B.v) The velocity vector dβ is continuous on [−α, α]\{0}. Morover, limθ0 dβ (θ )
and limθ0 dβ (θ ) exist, and the supremum in (15) is taken over [−α, α]\{0}.
Remark 4 (i) The original definition of spiraling curves [21, Sect. 3.3] only assumed
that there exists at least one angle such that the conditions (A.i) − (A.v) are
satisfied. Let us mention here that our additional assumptions are also met by
the examples mentioned in [21]. In particular, the set of concentric circles is also
a spiraling curve in the sense of Definition 3.
(ii) Since the collection of star shaped polygons and paths (as defined in Theorem 1)
generated by a set of points consist of countably many line segments (mostly
parallel and equispaces within escape cones), it is a straightforward task to show
that these trajectory sets are indeed spiraling curves. The only escape cones that
need more attention are those intersecting the line segments s(kz N , (k + 1)z 1 ).
In the limit however, these are parallel and equispaced line segments.
Subsequently, we give a full characterization of the set of all weak limits of translates
of spiraling curves. To do so, we establish two technical lemmas that describe the weak
limits of z k + Γ according to a certain geometric condition on the sequence {z k }k∈N .
Lemma 2 Let z k = −rk (θk ), k ∈ N, be an unbounded sequence. If there exist no

pair (N , γ ), N ⊂ N, γ ∈ IΓ , such that
(i) {rk }k∈N is unbounded and (ii) {rk sin(2π(θk − γ ))}k∈N is bounded,
then there exist a subsequence {z kn }n∈N , an angle β ∈ T\IΓ , and a constant τ ∈ R,

such that (a) −z kn ∈ Int(Cα,β ), for every n ∈ N, (b) θkn → θ ∗ , n → ∞, and
w
(c) z kn + Γ → τ d⊥ + L d,λZ ,

where d = limn→∞ dβ (β − θkn ) and λ = ηβ (β − θ ∗ ) sin arccos (θ ∗ ) · d .
Moreover, for every τ ∈ R, β ∈ T\IΓ and for d = limθ0 dβ (θ ) as well as for
d = limθ0 dβ (θ ), there exist a sequence {z k }k∈N such that
w
z k + Γ → τ d⊥ + L d,λZ ,

with λ = ηβ (0) sin arccos (β) · d .
Remark 5 The distinction of the two directions d = limθ0 dβ (θ ) and d =

limθ0 dβ (θ ) (which coincide for β ∈ T\IΓ ) in Lemma 2 is necessary as, for β ∈ IΓ ,
the sets of parallel lines with the left and right limits of the asymptotic velocity as direc-
tions are both included in WΓ .
Proof Step 1 Since rk sin(2π(θk − γ )) is unbounded for every γ ∈ IΓ and since T is

compact we can, after passing to a subsequence, assume that {−z k }k∈N is contained
in the interior of an escape cone Cα,β for some β ∈ T\IΓ and that θk → θ ∗ ∈
[−α + β, α + β].
As the problem is rotation invariant, we may for simplicity assume that β = 0. In
the following we therefore omit the subscripts in ηβ , rβ , etc.. As z k is unbounded,
we can, after passing to yet another subsequence, write z k as

z k = − η(θ ∗ )(n k + vk ) + ρ(θ ∗ ) (θk ),
where n k ∈ N is strictly increasing, vk ∈ [0, 1] converges to v ∗ , and η, ρ are the

functions given by (A.iv). If z k + Γ converges weakly to Γ , then z k + vk η(θ ∗ ) (θk ) +
Γ converges weakly to v ∗ η(θ ∗ ) (θ ∗ ) + Γ . As z k + vk η(θ ∗ ) (θk ) still satisfies the
assumption of this lemma, we may further simplify notation and assume
z k = −(η(θ ∗ )k + ρ(θ ∗ )) (θk ).
Step 2 We will later (in Step 4) show that the following property is always satisfied:
For every compact set K ⊂ R2 there exists k ∗ ∈ N such that
K ⊂ z k + Cα−|θk |,θk , k ≥ k ∗ .
For h ∈ Cc (R2 ), there exist Tθ , Sθ > 0 and tθ ∈ R such that the parallelogram

Pθ = (tθ + t) (θ ) + s d(θ ) : s ∈ [−Sθ , Sθ ], t ∈ [0, Tθ ]
satisfies supp(h) ⊂ Pθ . Note that Tθ , Sθ can be chosen to depend continuously on θ ∈

[−α, α]. In order to simplify the argument, we define a slightly larger parallelogram
Pθ containing Pθ . To this end, let m θ , Mθ ∈ Z be the largest (respectively smallest)
integer such that m θ η(θ ) ≤ tθ , and Mθ η(θ ) ≥ tθ + Tθ and define

Pθ = (m θ − 1/2 + t)η(θ ) (θ ) + s d(θ ) : s ∈ [−Sθ , Sθ ], t ∈ [0, Mθ − m θ + 1] ,
as well as the smaller parallelograms

Pθ,m = (m − 1/2 + t)η(θ) (θ ) + s d(θ) : s ∈ [−Sθ , Sθ ], t ∈ [0, 1] , m θ ≤ m ≤ Mθ .
Since all parameters defining Pθ are continuously depending on θ , it first follows that
the number smaller parallelograms is bounded by maxθ∈[−α,α] Tθ /η(θ ) + 2. Secondly,
there exists a compact set K such that Pθ ⊂ K for ever θ ∈ [−α, α]. Consequently, by
our previous assumption, for k large enough we have that Pθk ⊂ K ⊂ z k + Cα−|θk |,θk .
Step 3 For each k, m, let ψk,m : Ik,m → R2 be a re-parametrization by arc-length of
the segment {z k + γ (k + m + θ )}θ∈[−α,α] such that

ψk,m (0) = z k + γ (k + m + θk ) = r (k + m + θk ) − kη(θ ∗ ) − ρ(θ ∗ ) (θk ).
For large values of k, ψk,m (0) approximates mη(θ ∗ ) (θ ∗ ). Applying a first order
Taylor approximation yields

ψk,m (t) − r (k + m + θk ) − kη(θ ∗ ) − ρ(θ ∗ ) (θk ) − tψ (0)
k,m
t2
≤ sup ψ (s). (16)
2 s∈Ik,m k,m
Let us define Smax := maxθ∈[−α,α] Sθ (and Smin analogously). By conditions (A.iii)-

(A.v) of Definition 3 and

sup ψk,m (s) = sup |κ(k + m + θ )| → 0, as k → ∞,
s∈Ik,m θ∈[−α,α]
we have that for every δ > 0 there exists k ∗ ∈ N such that

(i) |κ(k + m + θ )| < δ/8Smax2 ,

γ (k+m+θk )
(ii) ψk,m (0) − d(θk ) = γ (k+m+θk ) − d(θk ) < δ/8Smax ,
(iii) |r (k + m + θ ∗ ) − η(θ ∗ )(k + m) − ρ(θ ∗ )| < δ/4,
(iv) |r (k + m + θk ) − r (k + m + θ ∗ )| < δ/4,
whenever k ≥ k ∗ . Note that we have used that the C 2 -norm of r is globally bounded
to ensure the last estimate. Now, if t ∈ [−2Smax , 2Smax ] and k ≥ k ∗ , then, using (16)
and triangle inequality, we get

ψk,m (t) − η(θ ∗ )m (θk ) − t d(θk )
≤ |r (k + m + θ ∗ ) − r (k + m + θk )|
+ |r (k + m + θ ∗ ) − η(θ ∗ )(k + m) − ρ(θ ∗ )|
t2
+ |t|ψk,m (0) − d(θk ) + sup ψ (s)
2 s∈Ik,m k,m
< δ/4 + δ/4 + δ/4 + δ/4 = δ.
Hence, if δ < min{Smin , η}, then ψk,m (Ik,m ) ∩ Pθk ⊂ Pθk ,m , and ψk,m (t) ∈
/ Pθk ,m for
every |t| ≥ 2Smax .
As h is uniformly continuous we may choose δ according to ε such
that x − y < δ
implies |h(x)−h(y)| < ε. Let λk = η(θk ) sin arccos( (θk )· d(θk )) . Then, for k ≥ k ∗

h(x) dH (x) −
1
h(x) dH (x)
1
z k +Γ L d(θ ),λ Z
k k
Mθk

= h(x) dH 1
(x) − h(x) dH (x)
1
(z k +Γ )∩Pθ ,m
m=m θk k L d(θk ),λk Z ∩Pθk ,m
Mθk
2Smax

≤ h(ψk,m (t)) − h η(θk )m (θk ) + d(θk )t dt
m=m θk −2Smax

M
θk

≤ 4Smax sup h(ψk,m (t)) − h η(θk )m (θk ) + d(θk )t
m=m θk t∈[−2Smax ,2Smax ]

≤ 4Smax (Mθk − m θk )ε ≤ 4Smax max Tθ /η(θ ) + 2 ε.
θ∈[−α,α]
As θk → θ ∗ and consequently λk → λ∗ , it is clearly possible to find k ∗ such that

h(x) dH (x) −
1
h(x) dH (x) ≤ ε,
1
L L d(θ
d(θ ∗ ),λ∗ Z k ),λk Z
w
whenever k ≥ k ∗ . Therefore, by triangle inequality, the convergence z k + Γ →
L d(θ ∗ ),λ∗ Z follows.
Step 4 It remains to show that the assumption K ⊂ z k + Cα−|θk |,θk can always be
satisfied for k large enough. To this end, we distinguish the cases |θ ∗ | < α and
|θ ∗ | = α. In the former case, we set α ∗ = (α − |θ ∗ |)/2 and observe that for k large
enough − 21 z k (θ ∗ ) + Cα ∗ ,θ ∗ ⊂ z k + Cα−|θk |,θk and the left side of the inclusion
relation will eventually cover any compactum.
If |θ ∗ | = α and θ ∗ ∈
/ IΓ , then it is possible to apply a rotation by a small angle
such that α is still the opening angle for the escape cone of β = 0 and |θ ∗ | < α.
Fig. 2 The set from the left hand

side of (17) (gray shaded area)
and the cone z k + Cθk ,θk
zk + Cθk ,θk
zk θk
If |θ ∗ | = α and θ ∗ ∈ IΓ , then one can rotate the problem so that θ ∗ = 0. Moreover,

we can without loss of generality assume that θk > 0. Let us consider the cone Cθk ,θk .
A simple geometric argument shows that

(x, y) ∈ R2 : x ≥ −ak , |y| ≤ bk , and 2bk x + ak bk ≥ ak y ⊂ z k + Cθk ,θk ,
(17)
where ak = rk cos(2π θk ) and bk = rk sin(2π θk ), see Fig. 2 for an illustration.

The assumptions that there is no unbounded subsequence of {z k }k such that
rkn sin(2π θkn ) is bounded and |θk | ≤ α ≤ 1/8 then shows, after possibly passing
to yet another subsequence, that z k + Cθk ,θk eventually covers any compactum.
Step 5 To see that every such collection of parallel lines is in fact a weak limit, we
may assume that d = limθ0 dβ (θ ) (d = limθ0 dβ (θ ) works exactly the same) and
choose
√
z k = τ d⊥ − (ηβ (0)k + ρβ (0)) β + 1/ k , β ∈ T\IΓ .
For this choice, the assumptions of the first part of this lemma are satisfied and one
w
may repeat the arguments to show that z k + Γ → τ d⊥ + L d,ηZ .

Lemma 3 Let z k = −rk (θk ), k ∈ N, be an unbounded sequence. If there exists a pair

(N , γ ), N ⊂ N, γ ∈ IΓ , such that
(i) {rk }k∈N is unbounded and (ii) {rk sin(2π(θk − γ ))}k∈N is bounded,
then there exist a subsequence {z kn }n∈N , and z ∈ R2 , such that

w
z kn + Γ → z + E γ ,
where the set of parallel edges E γ is given by

E γ := ηγ k (γ ) − t dγ− : k ∈ Z, t ∈ [0, ∞)

∪ ηγ k (γ ) + t dγ+ : k ∈ Z, t ∈ [0, ∞) ,
and ηγ = ηγ (0), dγ− = limθ0 dγ (θ ), and dγ+ = limθ0 dγ (θ ).

Moreover, for every γ ∈ IΓ and every z ∈ R2 there exist a sequence {z k }k∈N such
w
that z k + Γ → z + E γ .
Proof After potentially passing to a subsequence, one can assume that N = N, and that
rk is an increasing unbounded sequence. Moreover, by rotation invariance we may set
γ = 0. By assumption, the sequence rk sin(2π θk ) is bounded which shows that there
exists a subsequence converging to y ∗ . Therefore, after passing to this subsequence, we
see that if z k +Γ converges to Γ , then (−rk cos θk , 0)+Γ converges to (0, −y ∗ )+Γ .
Therefore, we further simplify the problem and assume that z k = (−rk , 0). From here
we can basically proceed as in the proof of Lemma 2 with some minor adjustments.
Let us shortly point out where caution is needed. For h ∈ Cc (R2 ) the parallelograms
P0 need to be replaced by arrow shaped objects defined as follows: let t ∗ ∈ R and
S, T > 0 be chosen such that the set

A0 := (t ∗ + t)e1 − s d0− : t ∈ [0, T ], s ∈ [0, S] ∪

(t ∗ + t)e1 + s d0+ : t ∈ [0, T ], s ∈ [0, S]
contains supp(h). Then one can proceed almost exactly as before by replacing P0 by
A0 , using the assumptions (B.iii) − (B.v) of Definition 3 and applying two Taylor
expansions for the left and right limits.
Again, we are left with showing that each such set of parallel edges is indeed a
weak limit of translates. Setting z k = z − (ηγ (0)k + ργ (0)) (γ ), γ ∈ IΓ , however
w
yields that z k + Γ → z + E γ .

Theorem 4 The set of weak limits of translates of a spiraling curve Γ is given by
WΓ = SΓ ∪ LΓ ∪ EΓ , (18)

where SΓ = {z + Γ : z ∈ R2 }, EΓ = z + E β : β ∈ IΓ , z ∈ R2 , and

LΓ = τ d⊥ + L d,λZ : τ ∈ R, d = lim dβ (θ ), or d = lim dβ (θ ), β ∈ T ,
θ0 θ0
and λ is defined as in Lemma 2.
Proof Lemmas 2 and 3 imply that SΓ ∪ LΓ ∪ EΓ ⊆ WΓ .

w
Now let z k + Γ → Γ . If z k ≤ C, then there exist a converging subsequence
z kn → z ∗ and it follows that Γ = z ∗ + Γ ∈ SΓ . If {z k }k∈N is unbounded, then either
there exist N ⊂ N and γ ∈ IΓ such that {z k }k∈N is unbounded and {rk sin(2π(θk −
γ ))}k∈N is bounded, or not. Hence, the assumption of either Lemma 2 or Lemma 3
are satisfied which leaves us with the limit Γ being either a set of parallel lines or a
set of parallel edges, i.e. WΓ ⊆ SΓ ∪ LΓ ∪ EΓ .

4.3 Spiraling Curves as Gabor Sampling Trajectories
After fully characterizing the set of all weak limits of translates we can now prove our
main results. We first show that, for a certain class of windows, spiraling curves are
Gabor sampling trajectories if and only if they are uniqueness sets for the weighted
p
modulation space M1/ϑs (R). Later, we then verify that certain spiraling curves are
indeed such uniqueness sets.
Theorem 5 Let Γ ∈ R2 be a spiraling curve, g ∈ span{h n : n ∈ N0 }, and 1 ≤ p < ∞.

Then Γ is a Gabor sampling trajectory for M p (R) if and only if Γ is a uniqueness
p
set for M1/ϑs (R) for some s > 2.
To show this theorem we first need to state two auxiliary lemmas. The first one is a
consequence of [16, Corollary 3.9(c) & Proposition 2.2].
Lemma 4 If there exist α > 0 such that |g(t)| e−α|t| and |
g (ξ )| e−α|ξ | , then
|Vg g(z)| 1/ϑs (z), for every s > 0.
The following lemma is a simple consequence of [6, Corollary 26].
p
Lemma 5 Let g ∈ span{h n : n ∈ N0 }. For every f ∈ M1/ϑs (R), 1 ≤ p < ∞, z ∈ R2 ,
and β ∈ T, it holds that Vg f (z + t (β)) is a real analytic function in t ∈ R.
Proof First, observe that Vg f (z + t (β)) is real analytic if and only if

Vg f R(β)R(−β)(z + t (β)) = VU (β)g (U (β) f ) R(−β)z + t e1
is real analytic. For β ∈ T one has that U (β)g ∈ M(R) by Lemma 4 and |U (β)g(t)|
e−α|t| . Since the metaplectic rotation leaves M1/ϑs (R) invariant, it follows by [6,
p

Corollary 26] that VU (β)g (U (β) f ) R(−β)z + t e1 is real analytic.

Proof of Theorem 5 First, observe that g ∈ M(R) by Lemma 4. We may there-

fore apply Theorem 3. By Theorem 4, the set of weak limits of translates consists
of finite shifts of Γ and the sets of parallel lines and edges. As Vg f (z − w) =
p
e2πiξ y Vg π(w) f (z), and M1/ϑs (R) is invariant under time-frequency shifts, it follows
p
that z + Γ is a uniqueness set for M1/ϑs (R) if and only if Γ is a uniqueness set for
p
M1/ϑs (R).
We now show that every set of parallel lines (resp. parallel edges) is a uniqueness
set. To show the two cases at once, we prove that any collection of parallel half lines
{ηk d1 + t d2 : t ≥ 0, k ∈ Z}, d1 , d2 not colinear, and η > 0, is a uniqueness set for
p
M1/ϑs (R). Since Vg f restricted to a line in R2 is real analytic by Lemma 5, it follows
that Vg f |{ηk d1 +t d2 : t≥0, k∈Z} = 0 implies Vg f |{ηk d1 +t d2 : t∈R, k∈Z} = 0.
The function U (θ )g is also a linear combination of Hermite functions and thus
has only finitely many zeros. Therefore, it follows that k∈Z supp(Tδk U (θ )g) = R
for every δ > 0. Finally, as g ∈ Aϑs (R), we may apply Proposition 3 to see that
Vg f |{ηk d1 +t d2 : t∈R, k∈Z} = 0 implies f = 0.

Corollary 2 Let g ∈ span{h n : n ∈ N0 }, 1 ≤ p < ∞, and η > 0. Moreover, let

P ⊂ R2 and {z 1 , ..., z n } ⊂ R2 be such that the assumptions (a) − (b) in Theorem 1
are satisfied. The collection of star shaped polygons Pη and the path S(z 1 , ..., z N ) are
Gabor sampling trajectories for M p (R).
Proof Take a line ⊂ R2 that contains one of the edges of Pη (resp. that contains the
line segment s(z 1 , z 2 )) and consider the collection {ηk }k∈N . If Vg f | Pη = 0 (resp.
Vg f |S (z 1 ,...,z n ) = 0), then, as Vg f is real analytic on any line ηk and zero on a
of nonzero measure, it follows that Vg f |{ηk }k∈N = 0. Arguing as before, we
subset
get k∈N supp(Tkη U (β)g) = R. Therefore, it follows that that f = 0.

Corollary 3 Let g ∈ span{h n : n ∈ N0 }, η > 0, and 1 ≤ p < ∞. The collection of

concentric circles Oη is a Gabor sampling trajectory for M p (R).
n
Proof For g = αk h k , we denote by F the polyanalytic function of order n
n k=0
given by F := k=0 αk Fk , where Fk (z) = Vh k f (z̄)eπ(z −z̄ )/4 eπ |z| /2 , see (6). Then
2 2 2
Vg f (z) = 0 exactly when F(z) = 0. Multiplying F by z n changes the zero set of F

at most in the origin, and results in a reduced polyanalytic function. By assumption,
and since Oη = Oη we hence have that z n F| Oη = 0. By Lemma 1, it thus follows
that F(z) = 0 for every z ∈ B R0 (0), where R0 can be chosen arbitrarily large.
Consequently, F(z) = 0 for every z ∈ C which implies that f = 0 and Oη is a
p
uniqueness set for M1/ϑs (R).

Acknowledgements The author would like to thank Felipe Negreira and José Luis Romero for valuable
discussions and suggesting some of the references. Moreover, we thank the anonymous referees for their
valuable input. Finally, we would like to acknowledge the support of the Austrian Science Fund (FWF)
through the Erwin Schrödinger Fellowship J-4254 and the project Y-1199.
Funding Open access funding provided by Austrian Science Fund (FWF).
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which
permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give
appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence,
and indicate if changes were made. The images or other third party material in this article are included
in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If
material is not included in the article’s Creative Commons licence and your intended use is not permitted
by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the
copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
References
1. Abreu, L.D., Bandeira, A.S.: Landau’s necessary density conditions for the Hankel transform. J. Funct.
Anal. 162, 1845–1866 (2012)
2. Abreu, L.D., Gröchenig, K.: Banach Gabor frames with Hermite functions: polyanalytic spaces from
the Heisenberg group. Appl. Anal. 91, 1981–1997 (2012)
3. Abreu, L.D., Speckbacher, M.: A planar large sieve and sparsity of time-frequency representations. In:
Proceedings of SampTA (2017)
4. Abreu, L.D., Speckbacher, M.: Donoho–Logan large sieve principles for modulation and polyanalytic
Fock spaces. Bull. Sci. Math. 171, 103032 (2021)
5. Ali, S.T., Antoine, J.-P., Gazeau, J.-P.: Continuous frames in Hilbert spaces. Ann. Phys. 222, 1–37
(1993)
6. Ascensi, G.: Sampling measures for the Gabor transform. J. Approx. Theory 200, 40–67 (2015)
7. Balk, M.B.: Polyanalytic functions and their generalizations. In: Gonchar, A.A., Havin, V.P., Nikolski,
N.K. (eds.) Complex Analysis I, pp. 197–253. Springer, Berlin (1997)
8. Benedetto, J.J., Wu, H.-C.: Nonuniform sampling and spiral MRI reconstruction. In: Proceedings of
SPIE Symposium on Wavelets Applications in Signal and Image Processing VIII (2000)
9. Bényi, A., Okoudjou, K.A.: Modulation Spaces. Applied and Numerical Harmonic Analysis.
Birkhäuser/Springer, New York (2020)
10. Beurling, A.: Local harmonic analysis with some applications to differential operators. In: Some Recent
Advances in the Basic Sciences. Belfer Graduate School of Science Annual Science Conference.
Yeshiva University, New York (1962)
11. Beurling, A.: On functions with a spectral gap. Seminar on Harmonic Analysis, University of Uppsala,
1942. In: Collected works of Arne Beurling: Vol. II. Harmonic Analysis. Birkhäuser, Boston (1989)
12. Christensen, O.: An Introduction to Frames and Riesz Bases. Applied and Numerical Harmonic Anal-
ysis. Birkhäuser, Boston (2003)
13. Folland, G.B.: Harmonic Analysis in Phase Space. Princeton University Press, Princeton (1989)
14. Fornasier, M., Gröchenig, K.: Intrinsic localization of frames. Constr. Approx. 22(3), 395 (2005)
15. Gröchenig, K.: Foundations of Time-Frequency Analysis. Applied and Numerical Harmonic Analysis.
Birkhäuser, Boston (2001)
16. Gröchenig, K., Zimmermann, G.: Spaces of testfunctions via the STFT. J. Funct. Spaces Appl. 2(1),
25–53 (2004)
17. Gröchenig, K., Romero, J.L., Unnikrishnan, J., Vetterli, M.: On minimal trajectories for mobile sam-
pling of bandlimited fields. Appl. Comput. Harmon. Anal. 39, 487–510 (2015)
18. Hartmann, A., Kamissoko, D., Konate, S., Orsoni, M.A.: Dominating sets in Bergman spaces and
sampling constants. J. Math. Anal. Appl. 495(2), 124755 (2021)
19. Jakobsen, M.S.: On a (no longer) new Segal algebra: A review of the Feichtinger algebra. J. Fourier
Anal. Appl. 24, 1579–1660 (2018)
20. Jaming, P., Speckbacher, M.: Planar sampling sets for the short-time Fourier transform. Constr. Approx.
53, 479–502 (2020)
21. Jaming, P., Negreira, F., Romero, J.L.: The Nyquist sampling rate for spiraling curves. Appl. Comput.
Harmon. Anal. 52, 198–230 (2020)
22. Jaye, B., Mitkowski, M.: A sufficient condition for mobile sampling in terms of surface density. Appl.
Comput. Harmon. Anal. 61, 57–74 (2022)
23. Kovrijkine, O.: Some results related to the Logvinenko–Sereda theorem. Proc. Am. Math. Soc. 129(10),
3037–3047 (2001)
24. Landau, H.J.: Necessary density conditions for sampling and interpolation of certain entire functions.
Acta Math. 117, 37–52 (1967)
25. Logvinenko, V.N., Sereda, Y.F.: Equivalent norms in spaces of entire functions of exponential type.
Teor. Funktsii Funktsional. Anal. i Prilozhen 19, 234–246 (1973)
26. Luecking, D.H.: Inequalities on Bergman spaces. Illinois J. Math. 25(1), 1–11 (1981)
27. Luecking, D.H.: Sampling measures for the Bergman space on the unit disc. Math. Annalen 316(4),
659–679 (2000)
28. Ortega-Cerdá, J.: Sampling measures. Publ. Mat. 42, 559–566 (1998)
29. Preiss, D.: Geometry of measures in Rn : distribution, rectifiability, and densities. Ann. Math. 125,
537–643 (1987)
30. Rahimi, A., Najati, A., Dehghan, Y.N.: Continuous frames in Hilbert spaces. Methods Funct. Anal.
Topol. 12(2), 170–182 (2006)
31. Ramanathan, J., Steger, T.: Incompleteness of sparse coherent states. Appl. Comput. Harmon. Anal.
2(2), 148–153 (1995)
32. Rashkovskii, A., Ulanovskii, A., Zlotnikov, I.: On 2-dimensional mobile sampling. Appl. Comput.
Harmon. Anal. 62, 1–23 (2023)
33. Seip, K.: Beurling type density theorems in the unit disk. Invent. Math. 113, 21–39 (1993)
34. Seip, K.: Interpolation and Sampling in Spaces of Analytic Functions. University Lecture Series, vol.
33. American Mathematical Society, Providence (2004)
35. Sun, W., Zhou, X.: Irregular Gabor frames and their stability. Proc. Am. Math. Soc. 131(9), 2883–2893
(2002)
36. Unnikrishnan, J., Vetterli, M.: Sampling and reconstruction on spatial fields using mobile sensors.
IEEE Trans. Signal Process. 61(9), 2338–2340 (2013)
37. Unnikrishnan, J., Vetterli, M.: Sampling high-dimensional band-limited fields on low dimensional
manifolds. IEEE Trans. Inform. Theory 59(4), 2103–2127 (2013)
38. Zayed, A.I.: Advances in Shannon’s Sampling Theory. CRC Press, Boca Raton (1993)
Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps
and institutional affiliations.

Sampling Trajectories For The Short-Time Fourier Transform: Michael Speckbacher

Uploaded by

Copyright:

Available Formats

Sampling Trajectories For The Short-Time Fourier Transform: Michael Speckbacher

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Sampling Trajectories For The Short-Time Fourier Transform: Michael Speckbacher

Uploaded by

Copyright:

Available Formats

Journal of Fourier Analysis and Applications (2022) 28:82

Sampling Trajectories for the Short-Time Fourier Transform

Received: 19 July 2021 / Revised: 22 September 2022 / Accepted: 5 October 2022

Keywords Gabor analysis · Sampling theory · Mobile sampling · Polyanalytic

Mathematics Subject Classification 42C15 · 42C30 · 42C40 · 65T60 · 94A20

Reconstructing a function from incomplete data is an omnipresent task in fields like

Communicated by Karlheinz Gröchenig.

1 Faculty of Mathematics, University of Vienna, Oskar-Morgenstern-Platz 1, 1090 Vienna, Austria

A more abstract form of the sampling problem is to characterize so called sampling

for some constants A, B > 0 and every F in a given subspace of L p (X ). A common

where h 0 denotes the standard Gaussian, and ϑ is a submultiplicative weight, see,

In most applications, Beurling’s theory is used to derive necessary conditions on

{ηz 1 , ηz 2 , ..., ηz N , 2ηz 1 , 2ηz 2 , ..., 2ηz N , 3ηz 1 , 3ηz 2 , ...}.

This paper is organized as follows. After a section dedicated to preliminaries, we study

2.2 Hausdorff Measures and Trajectory Sets

Let E ⊂ R2 . The 1-dimensional Hausdorff measure of E is given by

When restricted to a line, H1 equals the 1-dimensional Lebesgue measure.

μ(B R (z)) ≤ π ϕ(R)R.

A set E ⊂ R2 is called ϕ-regular if H1 | E is ϕ-regular.

2.3 The Short-Time Fourier Transform and Modulation Spaces

Let z = (x, ξ ) ∈ R2 . A time-frequency shift of a function g is given by

π(z)g(t) = Mξ Tx g(t) = e2πiξ t g(t − x).

Given a window g ∈ L 2 (R), the short-time Fourier transform of f ∈ L 2 (R) is defined

The short-time Fourier transform satisfies the following orthogonality relation

which implies that, if g2 = 1, Vg : L 2 (R) → L 2 (R2 ) is an isometry.

In particular, for f ∈ Aϑs (R) and g ∈ Mϑ1s (R) we have

f g M 1 ≤ f Aϑs g M 1 , (2)

2.4 Hermite Windows and Polyanalytic Functions

A function F : C → C is called polyanalytic of order n if it satisfies the higher order

where F0 , . . . , Fn : C → C are holomorphic functions.

is polyanalytic of order n [2], where we identify z = (x, ξ ) ∈ R2 with z = x +iξ ∈ C.

2.5 The Metaplectic Rotation

Clearly, U (θ ) is a unitary operator on L 2 (R) with U (θ )∗ = U (−θ ), and U (θ )h n =

As ϑs is radially symmetric, (8) implies that f M 1 = U (θ ) f M 1 , which allows

∞ (R), and g ∈ M 1 (R).

3 General Trajectory Sets

3.1 Relative Density

We call a trajectory set Γ ⊂ R2 (m, R)-dense if

inf H1 (Γ ∩ B R (z)) ≥ m > 0,

Proposition 1 (Ascensi) Let g ∈ M 1 (R) and 1 ≤ p < ∞. If Γ ⊂ R2 is a Gabor

sup H1 (Γ ∩ B R (z)) ≤ M. (9)

If Γ is a Gabor sampling trajecory for M p (R), then Γ is relatively dense.

3.2 Gaussian Window

where n(R, N , δ, z) is calculated by the following rule: take S R (z) := z +

m ≤ H1 (Γ ∩ B R (z)) ≤ M, for every z ∈ R2 . (10)

H1 (Γ ∩ B R (z)) ≥ H1 (Γ ∩ S R (z)) ≥ δn(R, N , δ, z) > δ R 2 .

m ≤ H1 (Γ ∩ B R (z)) ≤ H1 (Γ ∩ S2R (z))

Regrouping the terms yields and choosing δ = δ N = 1

which can be arbitrarily large as N → ∞. In particular, there exists N ∈ N such that

3.3 The Case of Parallel Lines

3.4 Connection to Discrete Sampling

Proposition 4 Let g, tg ∈ H 1 (R), R > 0 be chosen such that

and Γ ⊂ R2 be a trajectory set. If there exist m, M > 0 such that

0 < m ≤ H1 (Γ ∩ Q R (z)) ≤ M < ∞, for every z ∈ R2 , (13)

then for every f ∈ L 2 (R)

The lower sampling bound follows with a similar argument. 

4.1 Weak Limits

Definition 1 Let Γ , Γ  , Γn be trajectory sets in R2 . We say that {Γn }n∈N converges

The following characterization of Gabor sampling trajectories is an immediate conse-

The lower sampling bound follows with a similar argument.

Definition 1 Let Γ , Γ , Γn be trajectory sets in R2 . We say that {Γn }n∈N converges

Remark 5 The distinction of the two directions d = limθ0 dβ (θ ) and d =

and ηγ = ηγ (0), dγ− = limθ0 dγ (θ ), and dγ+ = limθ0 dγ (θ ).