Professional Documents
Culture Documents
net/publication/263661210
CITATION READS
1 431
2 authors:
Some of the authors of this publication are also working on these related projects:
The Quantum Arnold Transformation and the Ermakov-Lewis Invariant View project
All content following this page was uploaded by Manuel Calixto on 06 July 2014.
Chapter 10
Abstract
Classical Fourier analysis can be eventually regarded as a chapter in the representa-
tion theory of commutative groups. This group-theoretical perspective allows us to ex-
tend sampling theorems and discrete Fourier transforms to the non-Abelian case. For
this purpose, we use Coherent State or Wavelet techniques associated to unitary square
integrable (modulo a subgroup, possibly) representations of a Lie group G of motions
on a given manifold X (usually a homogeneous space G/H). Choosing a discrete
(finite or infinite) set of sampling points of X, we construct discrete frames and pro-
vide reconstruction formulas and discrete Fourier transforms for complex bandlimited
functions on X, through a sinc-type kernel, by using left- and right- Moore-Penrose
pseudo-inverses of frame (or sampling) operators. We also study the case where there
are not enough sampling points to completely reconstruct a given complex function on
X ( i.e., undersampling) but a partial reconstruction of it (an “alias”) is still possible
up to a given error.
1 Introduction
Harmonic Analysis is the branch of mathematics which studies the representation of func-
tions or signals as the superposition of basic waves (“harmonics”). It investigates and gen-
eralizes the notions of Fourier series and Fourier transforms. In the past two centuries, it has
become a vast subject with applications in areas as diverse as signal processing, quantum
∗ E-mail addresses: calixto@ugr.es, juguerre@um.es
312 M. Calixto and J. Guerrero
mechanics, number theory (viz, quadratic reciprocity law, Galois group, modular forms,
etc), ergodic theory (viz, equality of space and time averages in statistical mechanics), prob-
ability theory, the theory of automorphic functions, neuroscience, etc. It is perhaps with the
advent of the great electrical revolution when the applications of Fourier analysis to signal
processing achieved its greatest splendor. Filters, noise, discretization and sampling theo-
rems (like Nyquist-Shanon’s) constitute cornerstones in the fields of Telecommunications
and space exploration technologies.
Large sections of some of these subjects may be looked upon as nearly identical with
certain branches of the theory of group representations. They were Hermann Weyl and Fritz
Peter in 1927 who pointed out and emphasized the (still insufficiently appreciated) fact that
classical Fourier analysis can be illuminatingly regarded as a chapter in the representation
theory of compact commutative Lie groups. From this point of view, abstract harmonic
analysis can be also understood as the theory of group representations and their decompo-
sition into irreducibles. The theory for Abelian locally compact groups is called Pontryagin
duality and it is considered to be in a satisfactory state. Harmonic Analysis studies the prop-
erties of that duality and attempts to extend those features to different settings, for instance
to the case of non-Abelian Lie groups, where no general satisfactory theory is currently
known.
In the first part of this chapter, we shall review the Harmonic Analysis on a general
Abelian group G, as a natural extension of the standard Fourier Analysis on the real line
G = R. Useful concepts and definitions like the periodization ΠH and sampling ΞH oper-
ators, for a given discrete subgroup H of G, will be essential in the definition of sampling
theorems and discrete Fourier Transforms. See e.g. Figure 1 for the intertwining of Fourier
transforms and series in continuous and discrete time trough periodization and sampling
theorems acting on different representation spaces and (sub)groups. We address the inter-
ested reader to other Refs. like [1–3] for related expositions.
In the second part of this chapter, we shall address the study of sampling techniques
on non-Abelian groups and their homogeneous spaces. Our approach here will rely heav-
ily upon coherent states (see e.g. [4–6]), frames and discretizations. The notion of coherent
states is deeply rooted in quantum physics and its relationship to classical physics. The term
“coherent” itself originates in the current language of quantum optics (for instance, coher-
ent radiation). It was introduced in the 1960s by Glauber [7]. They were Aslaksen and
Klauder [8] who first studied the one-dimensional affine group, for the purely quantum me-
chanical purpose of generalizing the standard uncertainty relations “position-momentum”
(or time-frequency), for the Heisenberg group, to “dilation-translation”. Alex Grossmann
and collaborators [9], discovered the crucial link between the representations of the affine
group and an intriguing technique in signal analysis developed by the geophysicist Jean
Morlet, currently known as “wavelet analysis”. The same scheme can be extended to any
group of motions G, containing dilations, acting on a manifold X. In Section 3 we shall
report on this general framework discretization techniques. Finally, we shall conclude with
Section 4 devoted to some comments, useful references and applications.
Harmonic Analysis on Groups 313
G=R ΠZ G/Z = S1
Fourier Transform - Fourier Series
continuous time ΞZ⊥ continuous time
Ĝ = R d = Z⊥ = 2πZ
G/Z
? ?
Shannon Z ZN = Z/NZ
Sampling ΠNZ
Theorem Fourier Series - Fourier Transform
discrete time Ξ(NZ)⊥ discrete time
Ẑ = Ĝ/Z⊥ = S1 ẐN = ZN
FFT
{Rn , Zn , Tn , ZnN }
where
T = {eiθ , θ ∈ [0, 2π[}
denotes the 1-dimensional torus (the unit circle S1 , or the multiplicative unitary group U(1)
of complex numbers of modulus one, or the special orthogonal group SO(2) of rotations in
314 M. Calixto and J. Guerrero
two dimensions) and its discrete version, the finite or discretized circle:
2π
ZN ≡ Z/NZ = {ei N k , k = 1, 2 . . . , N − 1},
the multiplicative group of the N roots of unity or the additive group of remainders modulo
N. The elementary functions or basic waves (harmonics) of Fourier analysis over Abelian
groups are the so-called characters.
This group of characters is called the (Pontryagin) dual group of G and we denote it by Ĝ
Example 2.2. Consider G = R. All characters are given by the pure oscillations
χω (t) = eiωt , t ∈ R, ω ∈ R̂ = R.
For G = T we have
χn (θ) = einθ , θ ∈ [0, π[, n ∈ T̂ = Z.
Finally, the Pontryagin dual of G = ZN is the finite circle itself:
n
χn (k) = ei N k , k ∈ ZN , n ∈ ẐN = ZN .
where |G| stands for the total measure µG (G) of G. Arbitrary square integrable complex
functions over G, Ψ : G → C, Ψ ∈ L2 (G, dµG ), may be expanded into superpositions of
characters Z
Ψ(g) = Ψ̂(χ)χ(g)dµĜ (χ) (2)
Ĝ
Harmonic Analysis on Groups 315
where dµĜ stands for the usual (dual) Haar measure on Ĝ. The coefficients Ψ̂(χ) are square
integrable functions over the dual group Ĝ, i.e. Ψ̂ ∈ L2 (Ĝ, dµĜ ), and are called the Fourier
transform Ψ̂ ≡ FG Ψ of Ψ. They can be calculated by taking the scalar product with all
characters Z
FG Ψ(χ) = Ψ̂(χ) = hχ|ΨiG = Ψ(g)χ̄(g)dµG (g). (3)
G
The measure dµĜ can be normalized such that the following Plancherel-Parseval formula
holds: Z Z
|Ψ̂(χ)| dµĜ (χ) = |Ψ(g)|2 dµG (g),
2
(4)
Ĝ G
which states that the Fourier transform map FG : L2 (G) → L2 (Ĝ), Ψ 7→ Ψ̂ is an isometry
and hence “conserves energy”.
Example 2.5. The standard example is G = SO(2) ' S1 = T = U(1) with Haar measure
dθ/2π and Ĝ = Z, so that the harmonic analysis on G = SO(2) is nothing other than the
standard theory of Fourier series. In the same way, harmonic analysis on G = R and G = Z
gives the standard Fourier transform in continuous and discrete time, respectively. For
G = ZN we have Fourier series in discrete time (see Figure 1).
Both convolution products are associative, but, in general, non-commutative for non-
Abelian groups. Instead, we have always Ψ ∗L Φ = Φ ∗R Ψ. For Abelian groups, both
left- and right-convolution products coincide and acquire the familiar form:
Z
Ψ ∗ Φ(g0 ) = Ψ(g0 − g)Φ(g)dµG (g).
G
In Fourier space the convolution product is simply given by pointwise multiplication. More
precisely:
Theorem 2.6 (Convolution theorem). Let Ψ̂, Φ̂ ∈ L2 (Ĝ, dµĜ ) the Fourier transform of
Ψ, Φ ∈ L2 (G, dµG ), respectively. Then we have that:
\
FG (Ψ ∗ Φ)(χ) = Ψ ∗ Φ(χ) = Ψ̂(χ)Φ̂(χ), , ∀χ ∈ Ĝ.
Proof. By the isometry property (4) and the intertwining property (8) of the Fourier trans-
form we have
Z
Ψ ∗ Φ(g) = hUgL0 Ψ̃|ΦiG = hFGUgL0 Ψ̃|FG ΦiĜ = hÛgL0 FG Ψ̃|FG ΦiĜ = Ψ̂(χ)Φ̂(χ)χ(g)dµĜ .
Ĝ
By the inversion formula of the Fourier transform (2), the theorem is proved.
Ψ(h ◦ g) = Ψ(g), ∀h ∈ H,
Harmonic Analysis on Groups 317
that is, if Ψ is invariant (fixed point, UhL Ψ = Ψ) under the action of H. This definition coin-
cides with the usual one of a periodic function with period T ∈ R, Ψ(t + nT ) = Ψ(t), ∀t ∈
R, ∀n ∈ Z, when G = R and H = Z. Because an H-periodic function depends only on clases,
ġ ≡ g mod H, we may identify the set of H-periodic functions with functions on the quotient
group G/H. For example, standard periodic functions on R, can be identified with func-
tions on the circle T = R/Z. There is a natural map from functions over the whole group G
to H-periodic functions. It is given by the periodization operator:
where dµG/H is the natural invariant measure on G/H inherited from G. Note that we can
write formally (in the sense of distributions) the periodization operator as a convolution
product ΠH Ψ(g) = δH ∗ Ψ(g) with δH (g) = ∑h∈H δ(g − h) the so called delta-comb. Note
that H ⊥ is exactly the set of H-periodic characters. Indeed, for χ ∈ H ⊥ , g ∈ G, h ∈ H we
have χ(g ◦ h) = χ(g)χ(h) = χ(g). Moreover, since the property χ(h) = 1, which defines
the annihilator H ⊥ in the equation (9), is preserved under multiplication and division of
characters, the set H ⊥ is actually a closed subgroup of Ĝ. We always have H ⊂ (H ⊥ )⊥ ,
and the reverse inclusion is true if H is itself a closed subgroup of G. There is a remarkable
duality between subgroups and quotient groups of a LCA group, as stated in the following
theorem, which extends the theorem 2.3.
Theorem 2.8 (Pontryagin duality). Let H be a closed subgroup of a LCA group G and
define
[ → H ⊥ and ι2 : Ĝ/H ⊥ → Ĥ
ι1 : G/H
by ι1 (η) = η ◦ π, ι2 (χH ⊥ ) = χ|H , where π : G → G/H is the canonical projection. Then ι1
and ι2 are isomorphisms of topological groups.†
Let us suppose that G/H is compact. In this case, we shall call H a lattice. Let us
denote by |G/H| the lattice size (the measure of a fundamental domain of H in G or just,
the “period”).
Lemma 2.9 (Weil’s formula). The periodization operator ΠH allows us to write the scalar
product of an H-periodic function Ψ with an arbitrary function Φ over G as a scalar product
over G/H. More precisely:
hΨ|ΦiG = hΨ|ΠH ΦiG/H , (11)
for Ψ H-periodic. More explicitly:
Z Z
G
Ψ̄(g)Φ(g)dµG (g) =
G/H
dµG/H (g)Ψ̄(g) ∑ Φ(g + h).
h∈H
S
Proof. The key is to cover G as a disjoint union of fundamental domains G = h∈H (h +
G/H) and use that Ψ is H-periodic. See e.g. [2] for the complete proof.
† Note that the surjectivity of ι2 yields a sort of Hahn-Banach theorem for LCA groups, in the sense that
every character of H extends to a character of G.
318 M. Calixto and J. Guerrero
Example 2.10. For G = R and H = Z we have G/H = T, and the Weil formula reads:
Z ∞ ∞ Z 2π(n+1)
hΨ|ΦiR =
−∞
Ψ̄(t)Φ(t)dt = ∑ Ψ̄(t)Φ(t)dt
n=−∞ 2πn
∞ Z 2π
= ∑ Ψ̄(t + 2πn)Φ(t + 2πn)dt
n=−∞ 0
Z 2π ∞
=
0
dt Ψ̄(t) ∑ Φ(t + 2πn) = hΨ|ΠH ΦiT .
n=−∞
In general, Ψ could be any function for which point values on H make sense. We could also
consider an imperfect sampling operator Ξ∆H associated to some filter ∆ ∈ L2 (G) as:
which coincides with ΞH for ∆ = δ (the Dirac delta). Note also that one could formally
write Ξ∆H Ψ = ΞH (∆˜ ∗ Ψ), which means that the imperfect sampling is obtained by sampling
a smoothed version of Ψ perfectly. We are now ready to state the Poisson summation
formula, which links sampling to periodization via the Fourier transform.
Theorem 2.11 (Poisson summation formula). Let H ⊂ G be a lattice. On L1 (G, dµG ) ∩
FG−1 L1 (Ĝ, dµĜ ). We have:
ΞH ⊥ FG = FG/H ΠH , |G/H|FH ΞH = ΠH ⊥ FG . (14)
Proof. Let η ∈ H ⊥ . Then η is H-periodic. Using the Weil formula (11), we may write
(The second formula is dual of the first.) More explicitly, the Fourier expansion of ΠH Ψ in
L2 (G/H) is given by:
1
ΠH Ψ(g) = ∑ Ψ(g + h) = |G/H| ∑ hη|ΠH ΨiG/H η(g).
h∈H η∈H ⊥
Now, the Weil formula, hη|ΠH ΨiG/H = hη|ΨiG = Ψ̂(η), completes the proof.
Harmonic Analysis on Groups 319
ΠH ΞH
(G) - (G/H) (G) - (H)
FG FG/H FG |G/H|FH
? ? ? ?
ΞH ⊥ ΠH ⊥
(Ĝ) - (H ⊥ ) (Ĝ) - (Ĝ/H ⊥ )
Figure 2. Commutative diagrams defining the Poisson summation formula, where (G), (H),
etc, denote some suitable space of functions over G, H, etc, respectively.
Graphically, the Poisson summation formula (15) states that the diagrams of Figure 2
are commutative, where (G), (H), etc, denote some suitable space of functions over G, H,
etc, respectively.
Example 2.12. For example, for G = R and H = {t = nT, n ∈ Z}, for some period T ∈ R,
the Poisson summation formula states that, periodizing in the time representation (resp.
frequency representation) is equivalent to sampling in frequency space (resp. time rep-
resentation). More precisely, the following identity holds in the sense of a distributional
(Dirac delta) equality:
∞
2π ∞ 2πk
∑ −inT ω
e = ∑ δ ω− T .
T k=−∞
(16)
n=−∞
In general, neither the perfect nor the imperfect sampling operator has an inverse. How-
ever, there are subspaces on which the sampling operator may be inverted, that is, there are
functions on G which can be reconstructed from its samples on H ⊂ G. We shall give the
following definition
Definition 2.13. A subspace LH 2 (G) of L2 (G) is called a perfect H-sampling space if the
2 (G)
perfect sampling operator restricted to LH
ΞH : LH
2
(G) → L2 (H)
is an isomorphism and if, in addition, UhL ΞH = ΞH UhL for all h ∈ H ⊂ G. The same definition
applies to the imperfect sampling operator Ξ∆H : Ψ 7→ ΞH (∆˜ ∗ Ψ) with respect to some ∆ ∈
L2 (G).
The usual CWT on the real line R is derived from the natural unitary representation of
the affine group G = SIM(1)
t −b
[U(a, b)ψ](t) = a−1/2 ψ(t 0 ) ≡ ψa,b (t), t0 = (17)
a
in the Hilbert space L2 (R, dt) of finite energy signals ψ ∈ H . The same scheme applies to
the CWT on a general manifold X, subject to the transitive action, x → x0 = gx, g ∈ G, x ∈ X,
of some group G of motions on X (which can also contain dilations). If the measure dx
is G-invariant (i.e. d(gx) = dx), then the natural left action of G on the Hilbert space
(H , h·|·i) = L2 (X, dx) given by:
When dx is not strictly invariant (i.e. d(gx) = λ(g, x)dx), we have to introduce a multiplier
(Radon-Nikodym derivative of the quasi-invariant measure)
in order to keep unitarity (such is the case of the affine group). The fact that U(g2 )U(g1 ) =
U(g2 g1 ) (i.e., U is a representation of G) implies cohomology conditions for multipliers,
that is:
λ(g2 g1 , x) = λ(g2 , x)λ(g1 , g−1
2 x). (21)
Consider also the space L2 (G, dg) of square-integrable complex functions Ψ on G,
where dg = d(g0 g), ∀g0 ∈ G, stands for the left-invariant Haar measure, which defines the
scalar product Z
(Ψ|Φ) = Ψ̄(g)Φ(g)dg. (22)
G
1
dg = da db.
a2
A non-zero function γ ∈ H is called admissible (or a fiducial vector) if Γ(g) ≡ hU(g)γ|γi ∈
L2 (G, dg), that is, if
Z Z
cγ = Γ̄(g)Γ(g)dg = |hU(g)γ|γi|2 dg < ∞. (23)
G G
A unitary representation for which admissible vector exists is called square integrable.
For a square integrable representation, besides Eq. (23) the following property holds
(see [9]): Z
|hU(g)γ|ψi|2 dg < ∞ , ∀ψ ∈ H . (24)
G
322 M. Calixto and J. Guerrero
Let us assume that the representation U is irreducible, and that there exists a function γ
admissible, then a system of Coherent States (CS) of H associated to (or indexed by) G is
defined as the set of functions in the orbit of γ under G
γg = U(g)γ, g ∈ G. (25)
There are representations without admissible vectors, since the integration with respect
to some subgroup diverges. In this case, or even for convenience when admissible vectors
exist, we can restrict ourselves to a suitable homogeneous space Q = G/H, for some closed
subgroup H. Then, the non-zero function γ is said to be admissible mod(H, σ) (with σ : Q →
G a Borel section‡ ), and the representation U square integrable mod(H, σ), if the condition
Z
|hU(σ(q))γ|ψi|2 dq < ∞, ∀ψ ∈ H (26)
Q
Tγ : H −→ L2 (Q, dq)
hγσ(q) |ψi (28)
ψ 7−→ Ψγ (q) = [Tγ ψ](q) = √
cγ .
Its range Lγ2 (Q, dq) ≡ Tγ (H ) is complete with respect to the scalar product (Φ|Ψ)γ ≡
Φ|Tγ A−1 −1
σ Tγ Ψ Q and Tγ is unitary from H onto Lγ (Q, dq). Thus, the inverse map Tγ
2 −1
H
‡A section ρ : Q → G of the fibre bundle G → Q with base Q and fibre H is said to be a Borel section if it is
measurable with respect to the Borel σ-algebras of Q and G.
§ In this paper we shall extensively use the Dirac notation in terms of “bra” and “kets” (see e.g. [6, 18]).
The Dirac notation is justified by the Riesz Representation Theorem, and is valid in more general settings than
Hilbert spaces of square integrable functions .
Harmonic Analysis on Groups 323
which expands ψ in terms of CS A−1 σ γσ(q) with coefficients Ψγ (q) = [Tγ ψ](q). These formu-
las acquire a simpler form when Aσ is a multiple of the identity, as is for the case considered
in this article. R
When it comes to numerical calculations, the integral Aσ = Q |γσ(q) ihγσ(q) |dq has to
be discretized, which means to restrict ourself to a discrete subset Q ⊂ Q. The question is
whether this restriction will imply a loss of information, that is, whether the set S = {|qk i ≡
|γσ(qk ) i, qk ∈ Q } constitutes a discrete frame itself, with resolution operator
The operator A need not coincide with the original Aσ . In fact, a continuous tight frame
might contain discrete non-tight frames.
Let us assume that S generates a discrete frame, that is, there are two positive constants
0 < b < B < ∞ (frame bounds) such that the admissibility condition
bI ≤ T ∗ T ≤ BI, (32)
where I denotes the identity operator in H . This implies that A is invertible. If we define
the dual frame {|q̃i ≡ A −1 |qi}, one can easily prove that the expansion (reconstruction
formula)
|ψi = ∑ Ψk |q̃k i (33)
qk ∈Q
fulfilling Sk (ql ) = Plk . A projector is obtained, instead of the identity, to account for the fact
that an arbitrary set of overcomplete data {Ψk } ∈ `2 , can be incompatible with |ψi ∈ H , and
therefore they are previously projected (note that an overdetermined system of equations is
being solved).
This case will be named oversampling, since there are more data than unknowns (see
[21]). In other contexts, when eq. (31) holds, the set Q is said to be sampling for the space
H [13].
We shall be mainly interested in cases where there are not enough points to completely
reconstruct a given function ψ, i.e., undersampling, but a partial reconstruction is still pos-
sible. In these cases S does not generate a discrete frame, and the resolution operator A
would not be invertible. But we can construct another operator from T , B = T T ∗ , acting
on `2 .
The matrix elements of B are
Bkl = hqk |ql i , (37)
therefore B is the discrete reproducing kernel operator. If the set S is linearly independent,
the operator B will be invertible and a (right) pseudoinverse can be constructed for T ,
Tr+ ≡ T ∗ (T T ∗ )−1 , in such a way that T Tr+ = I`2 . As in the previous case there is another
operator, PS = Tr+ T acting on H which is an orthogonal projector onto the subspace H S
spanned by S . A pseudo-dual frame can be defined as
fulfilling Lk (ql ) = δkl . The alias ψ̂ is the orthogonal projection of ψ onto the subspace H S ,
that is, |ψ̂i = PS |ψi. The relative (normalized) distance from the exact ψ to the recon-
structed function ψ̂ is given by the relative error function:
s
kψ − ψ̂k hψ|I − PS |ψi
Eψ (H S ) = = (42)
kψk hψ|ψi
As mentioned above, we shall denote this case by undersampling, since there are not
enough data to fully reconstruct ψ. In other contexts, a set Q is said to be interpolating if,
for an arbitrary set of data {Ψk } there exists a |ψi ∈ H such that hqk |ψi = Ψk [13]. This
condition is satisfied in this case since Lk (ql ) = δkl .
Harmonic Analysis on Groups 325
where
0 1 ... 0
.. .. . . ..
. . . .
Π=
.. ..
(ΠN = IN , Πt = Π∗ = Π−1 = ΠN−1 ),
. . ... 1
1 0 ... 0
is the generating matrix of the circulant matrices and Pc (t) is the representative polynomial
of the circulant (we put Π0 ≡ IN ). This structure allows for a straightforward diagonalization
and inversion. Indeed, according to the general theory (see e.g. [22]), every circulant matrix
is diagonalizable, with a factorization B = FN DFN∗ given in terms of the discrete Fourier
matrix
1
(FN )nm = √ e−i2πnm/N , m, n = 0, 1, . . . , N − 1 . (44)
N
and the diagonal matrix D = diag(λ̂0 , . . . , λ̂N−1 ), whose eigenvalues λk can be computed
through its representative polynomial as
N−1
λk = Pc (z̄k ) = ∑ Cl z̄lk , zk = e2πik/N , k = 0, . . . , N − 1. (45)
l=0
326 M. Calixto and J. Guerrero
superconductors and superfluids (like Bose-Einstein condensates) are coherent states. Like-
wise, the Lowest Landau Level (LLL) wavefunctions in Quantum Hall Effect (characterized
by a quantization of the Hall conductance in two-dimensional electron systems subjected
to low temperatures and strong magnetic fields) are coherent states; the formulation of such
interesting effect on the hyperboloid SU(1, 1)/U(1) has been recently considered in [29]
(see also references therein for the extension to other geometries) and we believe that the
construction of discrete frames and sampling theorems on homogeneous spaces X = G/H
can be useful when considering numerical simulations of these systems. Here the dis-
cretization process turns out to be essential for computational applications. We hope that
the establishment of new sampling theorems for harmonic analysis on non-Abelian groups
and their homogeneous spaces will be of importance for numerical study and simulation of
those physical systems bearing that symmetries. Actually, there are some important general
results about sampling and efficient computation of Fourier transforms for compact groups
(see e.g. [30, 31]). However, a comprehensive study of the non-compact case is far more
involved, although there is a quite well developed theory of sampling on Riemannian man-
ifolds (see Refs. from [32] to [33]) with reconstruction formulas for bandlimited functions
on homogeneous spaces. Other results in this direction have been obtained for specific
groups (see e.g. [6] for a survey). For instance, we would like to point out Ref. [34] for
the motion group and its engineering applications [35] (namely in robotics [36]), Ref. [37]
for discrete frames of the Poincaré group and Ref. [38] for the formulation of Wavelets on
the 3+1 dimensional conformal group SO(4, 2), the last two cases with important potential
applications to Relativity Theory.
Acknowledgements
Work partially supported by the Fundación Séneca, Spanish MICINN and Junta de An-
dalucı́a under projects [08816/PI/08, 08814/PI/08], FIS2008-06078-C03-01 and FQM219,
respectively.
References
[1] M. Holschneider: Wavelets: an analysis tool, Oxford University Press (1998)
[2] G.B. Folland, A Course in Abstract Harmonic Analysis, CRC Press, Boca Raton, FL
(1995).
[5] J.R. Klauder and Bo-Sture Skagerstam: Coherent States: Applications in Physics and
Mathematical Physics, World Scientific (1985).
[6] S.T. Ali, J-P. Antoine and J.P. Gazeau, Coherent States, Wavelets and Their General-
izations, Springer (2000)
328 M. Calixto and J. Guerrero
[7] R.J. Glauber, The quantum theory of optical coherence, Phys. Rev. 130 (1963) 2529.
[8] E.W. Aslaksen and J.R. Klauder, Unitary representations of the affine group, J. Math.
Phys. 9 (1968) 206.
[9] A. Grossmann, J. Morlet and T. Paul, Transforms associated to square integrable group
representations I. General results, J. Math. Phys. 26 (1985) 2473-2479.
[11] A. Selberg, Harmonic analysis and discontinuous groups in weakly symmetric Rie-
mannian spaces with applications to Dirichlet series. J. Indian Math. Soc. (N.S.) 20
(1956), 47-87.
A. Selberg, Proceedings of the International Colloquium on Zeta Functions, Bombay
(1956)
[14] J-P. Antoine, D. Rosca, P. Vandergheynst, Wavelet transform on manifolds: Old and
new approaches, Appl. Comput. Harmon. Anal. 28 (2010) 189-202
[15] H. Führ, Painless Gabor expansions on homogeneous manifolds, Appl. Comput. Har-
mon. Anal. 26 (2009) 200-211
[17] S.T. Ali, J-P. Antoine and J.P. Gazeau, Square integrability of group representations
on homogeneous spaces I. Reproducing triples and frames, Annales de l’ Institut H.
Poincaré 55 (1991) 829-856
[18] M. Calixto and J. Guerrero, Wavelet transform on the circle and the real line: a uni-
fied group-theoretical treatment, Applied and Computational Harmonic Analysis 21
(2006) 204-229.
[19] O. Christensen: An introduction to Frames and Riesz Bases, Birkhäuser, Biston (2003)
[21] M. Calixto, J. Guerrero and J.C. Sánchez-Monreal, Sampling Theorem and Discrete
Fourier Transform on the Riemann Sphere, Journal of Fourier Analysis and Applica-
tions 14 (2008), 538-567
[24] J-P. Antoine and P. Vandergheynst, Wavelets on the 2-sphere: a group-theoretical ap-
proach, Appl. Comput. Harmon. Anal. 7 (1999) 262-291.
[25] J-P. Antoine and P. Vandergheynst, Wavelets on the n-sphere and related manifolds, J.
Math. Phys. 39 (1998) 3987-4008.
[26] I. Bogdanova, P. Vandergheynst and J.P. Gazeau, Continuous wavelet transform on the
hyperboloid, Appl. Comput. Harmon. Anal. 23 (2007) 285-306
[27] M. Calixto, J. Guerrero and J.C. Sánchez-Monreal, Sampling Theorem and Discrete
Fourier Transform on the Hyperboloid, Journal of Fourier Analysis and Applications
17 (2011) 240-264
[28] E.P. Wigner: Group Theory and its Applications to the Quantum Mechanics of Atomic
Spectra, Academic Press, New York (1959).
[29] M. Daoud and A. Jellal, Quantum Hall Droplets on Disk and Effective Weiss-Zumino-
Witten Action for Edge States, International Journal of Geometric Methods in Modern
Physics 4 (2007), 1187-1204.
[31] D. Maslen, Sampling of functions and sections for compact groups, Modern Signal
Processing 46 (2003), 247-280.
[32] I. Pesenson, A sampling theorem on homogeneous manifolds, Trans. Am. Math. Soc.
352 (2000), 4257-4269
[34] A. Kyatkin and G.S. Chirikjian, Algorithms for fast convolutions on motion groups,
Applied and Computational Harmonic Analysis 9 (2000), 220-241
[36] A. Kyatkin and G.S. Chirikjian, Computation of robot configuration and workspaces
via the Fourier transform on the discrete motion-group, International Journal of
Robotics Research 18 (1999), 601-615
[37] J-P. Antoine and A.L. Hohouto, Discrete frames of Poincaré coherent states in 1+3
dimensions, Journal of Fourier Analysis and Applications 9 (2003), 141-173