On a new multivariate sampling paradigm and a

polyspline Shannon function

Ognyan Kounchev and Hermann Render
January 11, 2008

Abstract
In [9] and [12] we have introduced and studied a new paradigm for car-
dinal interpolation which is related to the theory of multivariate polysplines.
In the present paper we show that this is related to a new sampling para-
digm in the multivariate case, whereas we obtain a Shannon type function
S (x) and the following Shannon type formula:
1 Z
X j
f (r ) = S e r f ej d :
j= 1 Sn 1

This formula relies upon in…nitely many Shannon type formulas for the
exponential splines arising from the radial part of the polyharmonic op-
erator p for …xed p 1.
Acknowledgement. The …rst and the second author have been partially
supported by the Institutes partnership project with the Alexander von
Humboldt Foundation. The …rst has been partially sponsored by the
Greek-Bulgarian bilateral project BGr-17, and the second author by Grant
MTM2006-13000-C03-03 of the D.G.I. of Spain.

1 Introduction
The classical Shannon-Kotelnikov-Whittaker formula represents a function by
means of the following series
1
X sin (t jT )
f (t) = f (jT )
j= 1
(t jT )

where T = and f is a continuous function in L2 (R) having a Fourier transform

with support in [ ; ] :
There have been many generalizations of this formula, see [7], [3], and ref-
erences therein. There exists intimate relation between sampling theorems and
Wavelet Analysis which have been studied exhaustively in [22], see also the
references therein.

1
 In the present paper we will consider a multivariate generalization of the
Shannon theory which is based on a semi-orthogonal Wavelet Analysis using
polysplines, and which has been recently developed in [9], [10], [1], [12], [11]. In
the case of what we call ”spherical sampling”our approach provides a formula for
the recovery of a function f from a ”band-limited class P V0 ”by taking its values
over the spheres centered at the origin of Rn and having radii ej for all j 2 Z:
In the second case, of what we call ”parallel sampling” we recover a function f
from a ”band-limited class P V0 ” by taking its values on all hyperplanes in Rn
de…ned by fx1 = jg for j 2 Z:
An interesting feature of our results is that they use essentially all advances in
the one-dimensional sampling theorems for Riesz basis as developed by Gilbert
Walter in his book [22], see also the more recent paper [23].

2 Shannon-Walter sampling with exponential splines

(L splines)
In the present Section we will construct a formula of Shannon type which is
based on exponential splines (these splines are called sometimes L splines), cf.
[9].

2.1 Shannon sampling for Riesz basis according to Gilbert

Walter
First, we will provide the construction of Gilbert Walter of sampling and the
corresponding Shannon-Walter functions for semi-orthogonal Wavelet Analysis
generated by a scaling function from a Riesz basis; the proofs follow the proofs
outlined in [22] for the case of orthogonal scaling functions.
Let us assume that the real-valued function (t) is continuous and has
shifts f (t j)gj2Z which represent a Riesz basis for the Hilbert subspace V0
in L2 (R) ; i.e.
V0 := closL2 (R) f (t j) : j 2 Zg ;
and there exist two constants A; B > 0 such that for arbitrary constants cj
holds
2
1
X X1 X1
2 2
A jcj j cj (t j) B jcj j ;
j= 1 j= 1 j= 1
L2 (R)
following the tradition we call scaling function of V0 .
For every element f 2 V0 we have
1
X
f (t) = fj (t j) : (1)
j= 1

Assume that the space V0 has the property that for every f 2 V0 with
f (j) = 0 for j 2 Z

2
follows f 0: Then there should exist a basis f j (t)gj2Z of V0 such that a
formula of Shannon type holds, i.e. for every f 2 V0
1
X
f (t) = f (j) j (t) for t 2 R:
j= 1

The problem of …nding the proper conditions on and the basis f j (t)gj2Z has
been resolved by Gilbert Walter, [22]. Below we provide in detail his construc-
tion.
It is well-known that to the Riesz basis f (t j)gj2Z there exists unique
n o
dual Riesz basis e (t j) where duality means
j2Z
D E
(x j) ; e (x `) = j` ; (2)

cf. [21], [22]. Let us assume that the functions and e satisfy the following
asymptotic conditions:

(t) = O jtj
1 "
; e (t) = O jtj 1 "
(3)
as t ! 1; t 2 R;
or, which is equivalent, there exist positive constants C1 > 0 and C2 > 0 such
that for all t 2 R holds
C1
j (t)j 1+"
(1 + jtj)
e (t) C2
1+" :
(1 + jtj)
Then a direct estimate shows that the function
X
q (x; y) = e (x j) (y j) (4)
j2Z

is uniformly convergent on every compact set in R2 .

On the other hand for every …xed x1 2 R (or uniformly for x1 from a compact
subset in R ) we have the estimate
2
X C22
jq (x1 ; y)j j (y j) (y k)j 1+" 1+" :
j;k2Z
(1 + jx1 jj) (1 + jx1 kj)
We have also
Z Z
C1 C1
j (y j) (y k)j dy 1+" 1+" dy
R R (1 + jy jj) (1 + jy kj)
Z
1
C12 1+" dy
R(1 + jy jj)
Z
2 1
= C1 1+" dy < 1:
R (1 + jyj)

3
The last implies the convergence of the integral
Z
2
jq (x1 ; y)j dy:
R

Thus for every element f 2 L2 (R) the integrals

Z
q (x1 ; y) f (y) dy
R

make sense. In a similar way follows that q (x; y1 ) 2 L2 (R) uniformly for y1
from a compact subset in R:
Thus, from (1) and the representation for the dual basis we obtain the fol-
lowing reproduction property for every f 2 V0 ;
Z
q (x; y) f (y) dy = f (x) ; (5)
Z
q (x; y) f (x) dx = f (y) : (6)

Let us note that by the de…nition it is not clear whether the function q is
symmetric.
Let us remark that from (4) it is obvious that

q (t j; 0) = q (t; j) : (7)

We summarize the main results of Gilbert Walter’s approach from [22] in

the following Proposition.

Proposition 1 Let (t) be a scaling function satisfying (3), i.e. (t) and its
dual de…ned by (2) satisfy (t) ; e (t) = O jtj
1 "
as t ! 1 for t 2 R: Let
us de…ne the function by
X
( ) := (j) e i j : (8)
j2Z

We assume that the function ( ) satis…es the non-zero condition

( ) 6= 0 for 2 R: (9)

Then the following hold:

1. The system of functions fq (t; j)gj2Z = fq (t j; 0)gj2Z represents a
Riesz basis of V0 ; where q is de…ned in (4)
2. The unique dual Riesz basis fSj (t)gj2Z corresponding to fq (t j; 0)gj2Z
satis…es
Sj (t) = S0 (t j) :
Hence fS0 (t j)gj2Z is a Riesz basis as well.

4
 3. The following Shannon type formula holds for all f in the space V0 :
X
f (t) = f (j) S0 (t j) for t 2 R: (10)
j2Z

4. The function S0 (t) has a Fourier transform which satis…es

b( )
c0 ( ) =
S for 2 R: (11)
( )
Remark 2 We will call the function S0 Shannon-Walter function. Formula
(10) may be considered as the analog to the Shannon-Kotelnikov-Whittaker for-
mula whereby the condition for band-limitedness of f (i.e. the compactness of
the support of the Fourier transform fb( ) ) is replaced by f 2 V0 . In the context
of Wavelet Analysis the frequency variable of Fourier Analysis is replaced by
the index j.
Proof. 1. We will prove that the linear operator T which is de…ned by the
map h i
T e ( j) (x) = q (x; j)
c0
has a continuous extension to V0 : For this purpose we will consider the space V
of the Fourier transforms of the functions in V0 :
Let us consider the operator T1 de…ned on the space V c0 as the operator of
multiplication by ( ), i.e.
T1 [f ] ( ) := f ( ) ( ):
Since by (3) the function has a fast decay it follows that ( ) is a bounded,
continuous and periodic function and satis…es
j ( )j c0 > 0 for 2 R:
Hence, the linear operator T1 is bounded and its inverse de…ned by multiplication
1
with ( ) is also bounded.
Let us see that
1
T =F T1 F
where F denotes the Fourier transform and F 1 its inverse. Indeed, it is enough
to check this on the elements e (x j) : We have
h i
b
F e ( j) ( ) = e ij e ( )
h i
b
T1 F e ( j) ( ) = ( )e ij e ( )
!
X b
= (`) ei ` e ij e ( )
`2Z
b X
= e( ) (`) ei (` j)
;
`2Z

5
 h i
and taking F 1 shows …nally that F 1 T1 F e ( j) (t) = q (t j; 0) :
So we have proved that T is a bounded invertible operator, and by a basic
result in [21] (p. 30 ) it follows that the system fq (t; j)gj2Z is a Riesz basis of
V0 :
2. By a theorem in [21] there is a unique Riesz basis Sj (t) which is biorthog-
onal to fq (t j; 0)gj2Z ; i.e.

hq (t j; 0) ; Sk (t)i = jk :

On the other hand due to a change of variable = t k we have

Z
hq (t j; 0) ; S0 (t k)i = q (t j; 0) S0 (t k) dt
Z
= q ( + k j; 0) S0 ( ) d

= k j;0

which shows that the family of functions fS0 (t j)gj2Z is biorthogonal to

fq (t j; 0)gj2Z : Hence, by the uniqueness it follows Sj (t) = S0 (t j) :
3. Equation (10) is obtained by the expansion of f (t) in the basis fS0 (t j)gj2Z :
Indeed, from X
f (t) = fj S0 (t j)
j

and by the biorthogonality relation (6) and (7) follows

f (j) = hf; q (t j; 0)i = fj :

4. For proving (11) we need to apply the Shannon expansion (10) to the
function and to take the Fourier transform.

2.2 The exponential splines (L splines)

2.2.1 Preliminaries on exponential splines
At the beginning of Wavelet Analysis the splines played and important role (see
some history in [4] and [13]). In 1991 Chui and Wang have constructed the
compactly supported spline wavelets, cf. [2]. A characteristic feature of this
Wavelet Analysis is that it is semi-orthogonal. Further Wavelet Analysis using
exponential splines has been initiated in [6] and has been discussed in detail
in [9], [10]. We will refer to the monograph [9] for all details of the outline on
exponential splines following below.
The exponential splines are de…ned by means of ordinary di¤erential opera-
tors with constant coe¢ cients given by polynomials
N
Y
L (z) := (z j) ; (12)
j=1

6
where j are some real constants. For simplicity sake we will denote by the
non-ordered vector of the numbers j

:= [ 1; 2 ; :::; N] ; (13)

where some of the j ’s may repeat (the number of repetitions of j is the

multiplicity j of this j ); this notation is a convenient way to avoid every time
explicitly writing the multiplicities of the j ’s. We will have the operator
N
Y
d d d
L := L := j (14)
dt dt j=1
dt

and the space UN of analytic solutions de…ned by

d
UN = UN ( ) := f (t) : L f (t) = 0 for t 2 R (15)
dt
which is well known to be of dimension N , cf. [15]. The space UN is generated
by the exponents
ts e t for s = 0; 1; :::; 1
where 2 and is its multiplicity, where obviously we have
X
N= ;

and the sum is over all di¤ erent values in :

The classical polynomial case is obtained when j = 0 for all j = 1; 2; :::; N:
The basic objects which we will consider are the so-called cardinal expo-
d
nential splines (L splines ) generated by the operator L dt : Namely, we
de…ne the space S by putting

S := f : f (t) 2 C N 2
(R) \ C 1 (R n Z) ; fj(j;j+1) 2 UN ; (16)
d
i.e. they are piecewise analytic solutions to the equation L dt f (t) = 0:
We will consider the Wavelet Analysis generated by the B splines arising
d
from the operator L dt which are called sometimes T B splines (cf. [17] –
these are cardinal exponential splines having minimal support). As in the classi-
cal polynomial spline theory there exists up to a factor only one such T B spline,
i.e. a cardinal exponential spline QN [ ] (t) = QN (t) 2 S supported in the in-
terval [0; N ] and satisfying

QN (t) > 0 for t 2 (0; N ) ;

cf. [5], [2], [17].

It is convenient to de…ne QN (t) by means of its Fourier transform (cf. p.
274 in [9]), namely we have
QN
j=1 e
j
e i
d
QN ( ) = QN for 2 R: (17)
j=1 (i j)

7
 Many important quantities which we will need are de…ned by means of the
T B spline QN (t) :
In the case of the classical polynomial splines Schoenberg introduced the
so–called Euler-Frobenius polynomial. In order to de…ne a generalization of the
Euler-Frobenius polynomials introduced by Schoenberg we will consider
the function (see Corollary 13:24; p. 235 in [9], Micchelli [14])
Z
1 1 exz
AN 1 (x; ) := z
dz; (18)
2 i L (z) e
where the closed contour surrounds all points j 2 but excludes all zeros
xz
of the function eez . If we put
N
Y N
Y
r ( ) := e j
; s ( ) := e j
(19)
j=1 j=1

then we obtain the following representation (see Corollary 13:25 p. 235 in [9])

N 1 (x; ) = r ( ) AN 1 (x; ) (20)

where N 1 (x; ) is a polynomial of degree N 1 of : The so-called Euler-

Frobenius polynomial

N 1 ( ) := N 1 (0; ) (21)

has degree N 2; cf. Corollary 13:25 in [9].

We have the following important Proposition (cf. [14], [16], or Theorem
13:31 on p. 237 and Corollary 13:53 on p. 253 in [9]).

Proposition 3 The polynomial N 1 ( ) has exactly N 2 negative zeros.

By means of the T B spline we may de…ne the Euler-Schoenberg expo-

nential spline, (see p. 254 in [9] ):
1
X
j
N 1 (x; ) = QN (x j) : (22)
j= 1

Note that the word ”exponential”has nothing to do with the exponential meant
in the present paper, but it is related to the following easy to check property

N 1 (x + 1; ) = N 1 (x; ) : (23)

We have the following important (cf. p. 255; Proposition 13:55 and Theorem
13:56 on p. 256; Corollary 13:57 p. 256 in [9]).

Proposition 4 1. The Euler-Schoenberg exponential spline in (22) satis…es the

following relation for all x with 0 x 1;
N 1 P
( 1)
1 (x; ) = e (x; ) : (24)
j j
N N 1 N 1

8
 2. If the vector is symmetric, i.e. satis…es = ; then

N 1 N
N 1 ; = N 1 ;z for all z 2 C; (25)
2 z 2

and the polynomial N 1 (0; z) 6= 0 satis…es

N 1 (0; z) 6= 0 for all z 2 C with jzj = 1: (26)

In particular, we see that from (23) and (24) the following representation
holds:
N N
N 1 ; = 2
N 1 (0; ) (27)
2
0 1
N 1 N
X
= ( 1) 2 +1
exp @ j
A N 1 ( ):
j

We will use further the following basic

Lemma 5 Let the vector be symmetric, i.e. satis…es = : Then the

following holds

j N 1 ( 1)j N 1 ei j N 1 (1)j for 2 R:

Proof. By Proposition 3 the polynomial N 1 ( ) has precisely N 2 zeros

vj < 0; and by Proposition 4, 1) and 2) they satisfy

vj vN 1 j =1 for j = 1; 2; :::; N 2:

Hence, we have the representation

N
Y2
N 1 ( )=D ( vj ) ;
j=1

where D is a non-zero coe¢ cient. By the reality of the zeros vj < 0 it follows
N 2
N
Y2 Y
2
i i i i 1
N 1 e = jDj e vj = jDj e vj e
j=1 j=1
vj
N 2 N 2
2
Y
2
e i
vj Y
2
1 2vj cos + vj2
= jDj = jDj
j=1
jvj j j=1
jvj j
N 2
Y
2
1 2vj + vj2
jDj
j=1
jvj j
=j N 1 (1)j :

9
In a similar way we obtain
N 2
Y2
1 + 2vj + vj2
i
N 1 e jDj
j=1
jvj j
=j N 1 ( 1)j :

2.3 Checking the conditions of Gilbert Walter

Now we put for the scaling function

(t) = QN (t) : (28)

A fundamental result in [9] (Theorem 14:6 ) says that for arbitrary non-ordered
vector the system f (t j)gj2Z represents a Riesz basis for the space V0
de…ned by means of the cardinal splines S as

V0 = S \ L2 (R) ; (29)

the explicit values of the Riesz constants is found in [9].

According to (17) the Fourier transform of the basic T B spline satis…es the
asymptotic
N
!
Y e j
e i
1
b( ) = Q d N ( )= =O for ! 1;
i N
j=1 j j j

and it is clear that the derivatives of b ( ) satisfy similar asymptotic.

As we mentioned before formula (2), the basis f (t j)gj2Z has a dual Riesz
n o
basis generated by a unique function e, i.e. the set of functions e (t j)
j2Z
generates a Riesz basis of V0 : For the dual function e one has the Fourier trans-
form (see p. 356 in [9] )

be b( )
( ) = CS2 h i for 2R (30)
2N 1
e (N ; ei )
0 1
XN
1
with CS = exp @ j
A;
2 j=1

where e = [ ; ] , i.e. it is the symmetrized vector of :

Proposition 6 For arbitrary integer N 1 and arbitrary non-ordered vector

of order N; the functions and e de…ned by (28) and (30) satisfy condition
(3).

10
 Proof. The asymptotic of the function is clear since QN (t) is a com-
pactly supported function, in particular, for every integer m 0 the following
asymptotic holds:

1
(t) = O m for t ! 1:
jtj

Now from
h Lemma
i 5, applied to the vector e and to the corresponding func-
tion 2N 1 e N ; e (whereby we use essentially the symmetry of the vector
i
h i
e ), and from formula (27), it follows that the function 2N 1 e N ; ei is
bounded, namely
h i
j 2N 1 ( 1)j 2N 1
e N ; ei j 2N 1 (1)j for 2 R;

here the Euler polynomial 2N 1 ( ) corresponds to the vector e :

On the other hand, we have the following equality proved by integration by
parts:
Z 1
e (t) = 1 b
ei t e ( ) d
2 1
Z 1
1 1 d2 b
= 2
1 2
ei t e ( ) d
2 1 1 + t d
Z 1
1 1 d2 be
= 2
ei t 1 ( )d :
2 1+t 1 d 2
h i
Let us estimate the last integral. Since 2N 1 e (N ; ) is a polynomial in
h i
; it follows that the derivatives of 2N 1 e N ; ei with respect to have nice
asymptotic for ! 1: Now by the above remarks about the asymptotic of
the derivatives of the function b ( ) and the representation formula (30) follows
the inequality
e (t) 1 + t2 C for t 2 R:

This justi…es the asymptotic condition for e in (3).

For proving the asymptotic of and e we needed no restrictions on the

vector : However the non-zero condition (9) is more restrictive. In fact, G.
Walter has shown that the classical splines (this corresponds to the case of all
j = 0 in ) generate a simple and natural Shannon type formula only in
the case of odd degree cardinal splines, [22]. Similar is the situation with the
exponential splines. For that reason we will restrict our attention to the case

N = 2p: (31)

11
Proposition 7 1. We have the equality
X
( ) := (j) e i j = N 1 [ ] 0; ei for 2 R;
j2Z

hence the non-zero condition (9) is equivalent to the condition

N 1 [ ] (0; ) 6= 0 for j j = 1: (32)

2. For arbitrary symmetric non-ordered vector of order N = 2p and for

given by (28) condition (9) holds, i.e.

( ) 6= 0 for 2 R:

Proof. By the de…nition of N 1 in (22) and by (24) follows

X X
(j) j = ( j) j = N 1 [ ] (0; )
j2Z j2Z
0 1
N 1 N
X
( 1)
= N 1
exp @ j
A N 1 ( ):
j=1
h i
(Please note the di¤erence between N 1 [ ] and 2N 1 e since they corre-
spond to di¤erent vectors). Now Lemma 5 gives us the lower and upper bounds
of N 1 ei from which we obtain immediately (9), i.e.
X
(j) e i j 6= 0 for 2 R:
j2Z

Note that by means of Lemma 5 we have used essentially the symmetry of

the vector e and the fact that N = 2p: We will apply the above Proposition 7
to the case of non-ordered vectors = k = [ 1 ; :::; N ] given by

j = k for j = 1; :::; p
k := ; (33)
j =k for j = p + 1; :::; 2p

where k is some integer.

On the other hand there are other vectors which are important for our
multivariate theory, which are not symmetric but close to being symmetric,
and for which the non-zero condition (9) holds; they have been considered in
the paper [12] and called nearly symmetric there. For us the following vectors
= k = [ 1 ; :::; N ] will be important:

1+j = k + 2j for j = 0; :::; p 1;

k := (34)
1+p+j = n k + 2 + 2j for j = 0; :::; p 1;

here k is some integer.

We have the following analog to Proposition 7.

12
Proposition 8 Let = k be de…ned by (34). Then satis…es the non-zero
condition (9).

Proof. In the proof of Proposition 13 in the paper [12], we proved that

AN 1 (0; 1) 6= 0

for every k 0: On the other hand by Proposition 3 the polynomial N 1 ( )

can have only real zeros with < 0. Since by (20) and Proposition 4 we have
1
N 1 [ k ] (0; )= s AN 1 (0; )

it follows that the function N 1 [ k ] (0; ) has no zeros for j j = 1:

Hence, in the above cases we may apply the results of G. Walter as presented
in the previous section. By Proposition 1, 5) it follows that there exists a
Shannon-Walter exponential spline S0 2 V0 = S \ L2 (R) such that its
Fourier transform satis…es
b( ) b( )
c0 ( ) =
S = : (35)
( ) i
N 1 [ ] (0; e )

From the above Propositions we obtain the following:

Theorem 9 Let N = 2p and be symmetric non-ordered vector of order N; or

be a vector given by (34). Then for every cardinal exponential spline f 2 V0
there holds a Shannon type formula:
1
X
f (t) = S0 (t j) f (j) : (36)
j= 1

3 The multivariate case

Our main purpose is to …nd a multivariate generalization of the Shannon-type
formula (10). For that purpose we will use a Wavelet Analysis which is gen-
uinely multivariate. We will work in the framework of the polyharmonic Wavelet
Analysis introduced and studied in [9]. It uses piecewise polyharmonic functions
(polysplines) which generalize the piecewise polynomial functions (and splines)
in the one-dimensional spline theory.
We assume that the integer p 1 is …xed.
We will consider the solutions to the polyharmonic equation
p
u (x) = 0

in annular domains, i.e. in annuli Aa;b = fx 2 Rn : a < jxj < bg given by two
numbers a; b > 0: Here
X n
@2
=
j=1
@x2j

13
is the Laplace operator and p is its p-th iterate.
For every integer k 0 we will assume that we are given an orthonormal
basis of the spherical harmonics of homogeneity degree k; which we will denote
ak
as usually by fYk;` (x)g`=1 ; cf. [20], [9]. Thus we have
Z
Yk;`1 ( ) Yk;`2 ( ) d = `1 `2 ;
Sn 1

here Sn 1 is the unit sphere in Rn , d is the spherical area measure, and j is

the Kronecker symbol taking on 1 for j = 0 and 0 elsewhere. We will work with
the spherical variables x = r where r = jxj ; 2 Sn 1 : It is known that for
special functions of the form f (r) Yk;` ( ) one has (see [9], p. 152 )
p
(f (r) Yk;` ( )) = Lp(k) f (r) Yk;` ( )

where the operator L(k) is given by

@2 n 1 @ k (n + k 2)
L(k) := + :
@r2 r @r r2
On the other hand by the change v = log r we obtain another operator Mk;p
satisfying
d d
Lp(k) = e 2pv Mk;p
dr dv
where
pY1
Mk;p (z) := (z k 2j) (z + n + k 2 2j) ;
j=0

cf. [9], Theorem 10:34: Here it is important that unlike Lp(k) drd
the opera-
d
tor Mk;p dv has constant coe¢ cients. Thus we are within the framework of
the exponential polynomials and we may consider exponential splines for the
operator L = Mk;p de…ned by the vector = k = [ 1 ; :::; N ] given by (34).
The following representation for polyharmonic functions in annulus is a fun-
damental result (cf. Theorem 10:39 in [9]):

Proposition 10 Every function u 2 C 1 (Aa;b ) satisfying p

u (x) = 0 in the
annulus Aa;b permits the representation
X
u (x) = uk;` (r) Yk;` ( ) ; for x = r (37)
k;`

where uk;` (r) is a solution to

d
Lp(k) uk;` (r) = 0 for a < r < b:
dr

14
 Let us assume that the integer k 0 is …xed.
Further we consider the space of exponential splines generated by the op-
d
erator L = Mk;p dv : We de…ne the space of cardinal exponential splines and
d
the Wavelet Analysis corresponding to the operator L = Mk;p dv as in (29),
namely we put
d
Ve0k = fe 2 L2 (R) : Mk;p fe(v)j(j;j+1) = 0; and fe 2 C p 2
(R) : (38)
dv
Let us recall that in the case of odd dimension n (since there are no repeating
d
frequencies in of (34)) the analytic solutions of Mk;p dv f = 0 are linear
combinations of the exponentials
n op 1
e(k+2j)v ; e( n k+2+2j)v :
j=0

Respectively, by the change of the variable v = log r we obtain the space

n o
V0k := f : f (r) = fe(log r) ; for fe 2 Ve0 : (39)

Finally, we may apply Theorem 9, and we will obtain the Shannon-Walter

exponential spline given by (35) which we denote further by Se0 (v) : By (36)
(k)

Se0 (v) satis…es a Shannon-type formula, i.e. for every element f 2 Ve0 we
(k) (k)

have the equality

1
X
e
f (v) = fe(j) Se0 (v j) :
(k)
(40)
j= 1

By means of the change r = ev we have f (ev ) := fe(v) and we put

S0 (r) := Se0 (log r) :

(k) (k)

Hence, for every f 2 V0k we have the Shannon-type formula

1
X (k)
f (r) = f ej S0 re j
: (41)
j= 1

(k)
3.1 Asymptotic of S0 for k ! 1
d
It is essential to see what is the asymptotic of Se0 (v) and of Se0 ( ) as k ! 1:
(k) (k)

This will be important for the existence and the regularity of the multivariate
Shannon type function.
We assume as above that

(t) = QN [ ] (t) (42)

for the nonordered vector = k = [ 1 ; :::; N] given by (34) (or (33) below),
and that the non-zero condition (9) holds.

15
 By formula (35) we have

d cN ( )
Se0 ( ) =
(k)
i
: (43)
N 1 [ k ] (0; e )

On the other hand by formula (24) we have the representation

0 1
( 1)
N 1 X
N 1 [ k ] (0; ) = N 1
exp @ j
A N 1 ( ): (44)
j

(k)
Lemma 11 The Fourier transform of the Shannon-Walter function S0 and
its Fourier transform satisfy the following asymptotic:
1. uniformly for 2 R

d 1
Se0 ( ) = O
(k)
for k ! 1;
k

2. uniformly for all t 2 R

Se0 (t) = O (1)

(k)
for k ! 1:

Proof. By (42) it follows directly that for appropriate constants k0 0 and

C > 0 and for all 2 R and k k0 we have
QN i
j=1 e e epk
j
c
N ( ) = QN C p:
j=1 (i j) ( 2 + k2 )

Now we will take care of the denominator in (43). We will apply a subtle
estimate of Theorem 11 in the paper [12] (please note the conventions in the
paper, where the size of the vector is numbered through N + 1 and here we
use N ). It says that there exist two constants C1 ; C2 > 0 and an integer k0 1
such that for k k0 holds
C1 C2
jAN 1 (0; )j for j j = 1:
kN 1 kN 1
From relation (20) we obtain

N 1 ( ) = r ( ) AN 1 (0; )

and from (44) we obtain

1
N 1 [ k ] (0; )= s AN 1 (0; ) :

From (19) it follows that the asymptotic of the polynomial s ( ) for k !1

with = ei is
s ( ) ekp :

16
Hence, for k ! 1 and for appropriate constant C3 > 0 we have
epk
d ( 2 +k 2 )p 1
Se0 ( )
(k)
C3 epk
= :
k2p 1
k
2. By de…nition we have
Z 1
Se0 (t) =
(k) ck ( ) d ;
ei t S 0
1

hence by a direct estimate and a change of variable 0 = =k we obtain

Z 1
e(k) 2p 1 1
S0 (t) C3 k 2 pd
1 ( + k2 )
= C30 :

(k)
Obviously the results of Lemma 11 hold for the Shannon function S0 (r) :=
Se0
(k)
(log r) in the variable r:

3.2 The Shannon polyspline

We have seen that for the vector k de…ned by (34) the non-zero condition (32)
(k)
is satis…ed and we may de…ne the Shannon-Walter exponential splines S0 :
We will call the following function (actually it is understood in a distribu-
tional sense) Shannon polyspline kernel:
ak
1 X
X (k)
S0 (r; ; ) := S0 (r) Yk;` ( ) Yk;` ( ) (45)
k=0 `=1
X1
(k)
= S0 (r) Zk ( ; )
k=0
X1
Se0
(k)
= (log r) Zk ( ; ) ;
k=0

where Zk ( ; ) is the zonal harmonic of degree k, cf. [20]. Let us note that for
every …xed the function S0 (r; ; ) is a polyspline of r , and for every …xed
the function S0 (r; ; ) is a polyspline of r :
We will characterize the smoothness-convergence properties of this series in
terms of Sobolev spaces. Let us recall that the function h de…ned by the series
X
h (x) := hk;` Yk;` ( )
k;`

belongs to the Sobolev space on the sphere H s Sn 1

for a real number s if
and only if X s
1 + k 2 h2k;` < 1;
k;`

17
cf. section 7 and section 22 in [19].

Lemma 12 As a function of (or ) the Shannon polyspline S0 satis…es

S (r; ; ) 2 H s Sn 1
; S (r; ; ) 2 H s Sn 1

where
3
s= n+ "
2
for every number " > 0:

Proof. We know that for some constant C > 0 and for all k 0; and
` = 1; 2; :::; ak holds
n
jYk;` ( )j Ck 2 1 ;
see e.g. [18], [8]. Also, and one has the estimate

ak C1 k n 2
;

cf. e.g. [20].

(k)
Since the coe¢ cients of the series (45) are hk;` = S0 (r) Yk;` ( ) we have
the following inequalities
X X 1
X
s s s
1 + k2 h2k;` 1 + k2 C 2 kn 2
C2 1 + k2 k 2n 4

k;` k;` k=0

which proves the theorem.

3.3 The polyharmonic Wavelet Analysis

In [9] we have introduced a polyharmonic Wavelet Analysis by considering
the following basic space
p
P V0 := closL2 (Rn ) f (x) : f (x) = 0 on ej < r < ej+1 ; x=r ; f 2 C 2p 2
(Rn ) :

Let us note that the radii ej appear in a natural way since by the represen-
tation (37) and Proposition 10 every function u 2 P V0 has coe¢ cients uk;` (r)
satisfying
uk;` 2 V0k ;
where the spaces V0k are those de…ned in (39). Hence, we have the following
decomposition in the sense of L2 metric (cf. [9])
ak
1 M
M
P V0 = V0k
k=0 `=1

which is generated by the expansion (37).

18
 As we have seen above, the Shannon polyspline kernel (45) is a function in
some Sobolev space H s and might be a distribution in the case of a negative
exponent. Hence, the formulas which we will write will be understood in a
distributional sense. For two functions (distributions) f 2 H s Sn 1 and g 2
H s Sn 1 with expansions
X X
f( )= fk;` Yk;` ( ) ; g( ) = gk;` Yk;` ( ) for 2 Sn 1
k;` k;`

n 1
we have the scalar product on S de…ned by an integral understood in a
distributional sense
Z X
hf; gi := f ( )g( )d = fk;` gk;` :
Sn 1
k;`

Note that the in…nite sum which gives sense to the above integral is absolutely
convergent for f 2 H s Sn 1 and g 2 H s Sn 1 :
In our main Theorem below we will use the spaces P V0 to mimic the one-
dimensional cardinal spline space V0 ; and to formulate an analog to the Shan-
non type formula.
Theorem 13 For every function f 2 P V0 \ C (Rn ) the following Shannon
type formula holds:
X1 Z
r
f (r ) = S0 j ; ; f ej d ;
j= 1
e
Sn 1

where S0 (r; ; ) is the Shannon polyspline function de…ned in (45), and

the integral is understood in distributional sense.
Proof. Since by de…nition f 2 C (Rn ) it follows that for all j 2 Z we have
f ej 2 L2 ; hence the following L2 convergent expansion for every …xed r > 0
holds true
X ak
1 X ak
1 X
X
f (x) = fk;` (r) Yk;` ( ) = fek;` (log r) Yk;` ( ) ;
k=0 `=1 k=0 `=1

where x = r :
We obtain in distributional sense the following equalities, for v = log r;
!
X Z
1
r X1 1 X
X ak
(k)
j j j
S0 j ; ; f e d = S0 re fk` e Yk;` ( )
j= 1 n 1
e j= 1 k=0 `=1
S
0 1
X1 Xak X1
@ (k)
= S0 re j fk` ej Yk;` ( )A
k=0 `=1 j= 1
0 1
X ak
1 X 1
X
Se0
@ (k)
= (v j) fk` ej Yk;` ( )A :
k=0 `=1 j= 1

19
 On the other hand since for every k 0 we have fk` (ev ) 2 V0k , by the
Shannon formula in (40) we obtain
0 1
1 X
X ak X1 1 X
X ak
@ e (k)
S0 (v j) fk` e j A
Yk;` ( ) = fk` (ev ) Yk;` ( )
k=0 `=1 j= 1 k=0 `=1

= f (ev )
= f (r )

This ends the proof.

Let us remark that the sampling

ff ( ) : 2 Zn g

is the one used in the usual multivariate generalizations of the Shannon type
formula. In Theorem 13 we use though the sampling

fk;` ej : k 0; ` = 1; 2; :::; ak ; j 2 Z

which de…nes a new multivariate sampling paradigm.

4 The polysplines on parallel strips

We consider very brie‡y for every integer p 1 the Wavelet Analysis generated
by the polysplines on strips (with periodic conditions), and the corresponding
Shannon polyspline function and Shannon formula.
We will introduce some convenient notations:

x = (t; y) 2 Rn
t 2 R and y 2 Tn 1
;
n 1 n 1
T := [0; 2 ] :

The polyspline Wavelet Analysis on strips is de…ned by means of the following

space, cf. [9],
p
P V0 := closL2 (Rn ) f (t; y) : f (t; y) = 0 for t 2
= Z; f 2 C 2p 2
R Tn 1
;

these are functions which are piecewise polyharmonic on every strip (j; j + 1)
Tny 1 ; for all j 2 Z , and which are 2 periodic in the variables y: The functions
f 2 P V0 permit the representation
X
f (t; y) = f (t) eih ;yi ;
2Zn 1

20
where for every 2 Zn 1
the function f (t) is a solution to the equation
p
d2 2
j j f (t) = 0 for t 2
= Z;
dt2
2
X
2
j j = j:
j

In the present case we have the vector = k = [ 1 ; :::; N ] given by (33)

for k = j j. For this = k and for every integer p 1 we obtain the operator
p
d2
Lp(k) := k2 : (46)
dt2

Now as in (29) we have similar de…nition of the cardinal exponential spline

spaces, namely we put

d
V0k := f 2 L2 (R) : Lp(k) f (t)j(j;j+1) = 0; and f 2 C 2p 2
(R) :
dt

For the vector k de…ned in (33), by Proposition 7 condition (32) always

holds. Hence, for every k 0 there always exists the Shannon-Walter expo-
(k)
nential spline S0 de…ned in (35).

Remark 14 The case k = 0 corresponds to the classical cardinal splines of odd

degree considered by G. Walter in [22].

Since all basic lemmata have been proved above also for this type of vectors,
we obtain the following Shannon type formula:

Theorem 15 For every f 2 P V0 \ C (Rn ) holds the Shannon type formula

1 Z
X
0
f (t; y ) = S0 (t j; y; y 0 ) f (j; y) dy;
j= 1 Tn 1

where S0 (t; y; y 0 ) is the Shannon polyspline distribution de…ned by

X ( ) ih ;yi ih ;y 0 i
S0 (t; y; y 0 ) = S0 (t) e e :
2Zn 1

Again we may compare the usual multivariate sampling ff ( ) : 2 Zn g with

our sampling paradigm

f (j) : 2 Zn 1
;j 2 Z :

Finally, an idea for future research might be the error analysis of the Shannon
type formulas of Theorem 13 and Theorem 15.

21
References
[1] Aldaz, J. M., O. Kounchev, H. Render, On real-analytic recurrence re-
lations for cardinal exponential B-splines, J. Approx. Theory 145 (2007),
253–265.
[2] Chui, C. K., An Introduction to Wavelets, Academic Press, Boston, 1992.
[3] Ciaurri, J.L. Varona, A Whittaker-Shannon-Kotel’nikov sampling theorem
related to the Dunkl transform. Proc. Amer. Math. Soc. 135 (2007), 2939–
2947.

[4] Daubechies, I., Ten Lectures on Wavelets, Society for Industrial and Ap-
plied Mathematics, SIAM, Philadelphia, 1992.
[5] de Boor, C., A practical guide to splines, Springer, New York, 1978.
[6] de Boor, C., R. DeVore and A. Ron, On the construction of multivariate
(pre)wavelets, Constr. Approx. 9 (1993), 123–166.
[7] Higgins, J.R., Sampling theory in Fourier and signal analysis: Foundations,
Clarendon Press, 1996.
[8] Kalf, H., On the expansion of a function in terms of spherical harmonics in
arbitrary dimensions, Bull. Belg. Math. Soc. 2 (1995), 361–380.
[9] Kounchev, O. I., Multivariate Polysplines. Applications to Numerical and
Wavelet Analysis, Academic Press, Boston 2001.
[10] Kounchev, O. I., H. Render, Wavelet Analysis of cardinal L-splines and
construction of multivariate prewavelets, In: Proceedings “Tenth Interna-
tional Conference on Approximation Theory”, St. Louis, Missouri, March
26-29, 2001.
[11] Kounchev, O. I., H. Render, The Approximation order of Polysplines, Proc.
Amer. Math. Soc. 132 (2004), 455-461.
[12] Kounchev O., Render, H., Cardinal interpolation with polysplines on an-
nuli, Journal of Approximation Theory 137 (2005) 89–107.
[13] Mallat, S., A wavelet tour of signal processing, 2nd Edition, Academic
Press, Boston, 1999.
[14] Micchelli, C. A., Cardinal L-spline, in Studies in Spline Functions and Ap-
proximation Theory, S. Karlin, C. Micchelli, A. Pinkus, and I. Schoenberg
(eds.), Academic Press, New York, 1976, 203–250.
[15] Pontrjagin, L. S. , Ordinary Di¤erential Equations, Addison-Wesley, 1962.
[16] Schoenberg, I.J., Cardinal Spline Interpolation, SIAM, Philadelphia, PA,
1973.

22
[17] Schumaker, L. L., Spline Functions: Basic Theory, J. Wiley, New York,
1981.
[18] Seeley, R., Spherical Harmonics, Amer. Math. Monthly 73 (1966) 115–121.
[19] Shubin, M., Pseudodi¤erential Operators and Spectral Theory, Springer-
Verlag, Berlin-Heidelberg-New York, 1987:
[20] Stein, E. M. and G. Weiss, Introduction to Fourier Analysis on Euclidean
Spaces, Princeton University Press, Princeton, 1971.
[21] Young, R., Non-harmonic Fourier series, Academic Press, New York, 1980.

[22] Walter, G., Wavelets and Other Orthogonal Systems with Applications,
CRC Press, Boca Raton, FL, 1994.
[23] Walter, G., Wavelets and sampling, In: Modern Sampling Theory, J.
Benedetto et al. (eds.), Birkhäuser, Basel, 2001:

Corresponding author:
1. Ognyan Kounchev: Institute of Mathematics and Informatics, Bulgarian
Academy of Sciences, 8 Acad. G. Bonchev St., 1113 So…a, Bulgaria
[email protected]; [email protected]
2. Hermann Render: Departamento de Matemáticas y Computación, Uni-
versidad de La Rioja, Edi…cio Vives, Luis de Ulloa s/n., 26004 Logroño, Spain
[email protected]

23