0% found this document useful (0 votes)
55 views8 pages

IJEAS0303039

This document summarizes a method for reconstructing two unknown overlapping functions (U(z) and V(z)) based solely on their sum (W(z)=U(z)+V(z)). It makes the following key assumptions: 1. U(z) and V(z) are smooth, mutually linearly independent functions, overlapping across a finite carrier z ≤ L, with centers of symmetry zU and zV respectively. 2. U(z) and V(z) exhibit generalized symmetry, where the value of the function at -z is predetermined by the value at z through a coefficient of symmetry. 3. Measurement noise σξ is negligible compared to characteristic magnitudes of W(

Uploaded by

erpublication
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
55 views8 pages

IJEAS0303039

This document summarizes a method for reconstructing two unknown overlapping functions (U(z) and V(z)) based solely on their sum (W(z)=U(z)+V(z)). It makes the following key assumptions: 1. U(z) and V(z) are smooth, mutually linearly independent functions, overlapping across a finite carrier z ≤ L, with centers of symmetry zU and zV respectively. 2. U(z) and V(z) exhibit generalized symmetry, where the value of the function at -z is predetermined by the value at z through a coefficient of symmetry. 3. Measurement noise σξ is negligible compared to characteristic magnitudes of W(

Uploaded by

erpublication
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 8

International Journal of Engineering and Applied Sciences (IJEAS)

ISSN: 2394-3661, Volume-3, Issue-3, March 2016

About the Recognition and Reconstruction Two


Unknown Functions From Known Their Tandem
Vladimir V. Arabadzhi

on the carrier z L . We find the single function U ( z ) and


Abstract Paper is devoted to the area of processing of
physical signals and data or artificial perception. The incoming
single function V (z ) in Figure 1-a .
data are presented in the form of defined sum of two unknown
functions, spaced on finite carrier. Practical examples of the Then we do important assumption: W (z ) is presented by the
functions searched for can be presented: (a) by very closely tandem (all variables and numbers, used below, are assumed
spaced two atomic spectrums of radiation; (b) by the output dimensionless) of two smooth functions U ( z ) and V (z ) ,
signal of some telescope, scanning two point-like sources with which are overlapping each other at any point of the same
angle distance between each other much less, than the angle carrier z L . No any point z z * on above carrier where
width of main petal of telescope's directivity pattern. So the function U ( z*) 0 at V ( z*) 0 or V ( z*) 0 at U ( z*) 0 .
presented result can be classified as some case of the known
Itself determination W (z ) as "two functions" nothing do not
problem of superresolution. It is shown, that two functions of
signify. In realities possible to name such combination "one
tandem (mutually overlapping each other) can be reconstructed
function", or suppose, that W (z ) consists of tens of functions.
separately by using only three it is correct chosen numbers. The
To avoid the uncertainty, we must say the important thing:
class of functions (and their tandems), which are overlapping
each other and allows single reconstruction of both searched function U ( z ) has a centre of symmetry zU , and function
functions and also conditions of the correctness of the decision V (z ) has a centre of symmetry zV . In addition the searched
are formulated. The method of separation of two functions had
functions U , V are assumed mutually linearly independent,
been tested numerically on the variety of typical practical
and we must find single solution for tandem U , V on chosen
examples. The stability of suggested numerical procedure is
class of functions (see below in Section IV). Such problems
confirmed in the presence of uncorrelated errors of
can appear in investigations of closely spaced spectrums of
measurements and calculations at each point of tandem.
atomic radiation or in resolutions of astronomic sources by
radio antennas [2]-[4]. Below we suppose negligible weak
Index TermsAbout four key words or phrases in
alphabetical order, separated by commas. noise (see Figure 1-b)

AW , AU , AV , (1)
I. INTRODUCTION
Usually the problem reconstruction of signals is considered in where symbols AW >0, AU >0, AV >0 mean the
the presence of noise and interpreted as extraction of signal
characteristic magnitudes
U ( z ) covered by noise, so the measured signal can be of the
form W ( z ) U ( z ) ( z ) , where ( z ) - stationary noise with
1/ 2
mean-square value AW [ L L W 2 ( z )dz /(2 L )] ,
2 1/ 2
( z ) 1/ 2
AU [ L L U 2 ( z )dz /(2 L )] ,
correlation interval
1/ 2
L [1]. AV [ L L V 2 ( z )dz /(2 L )]

Below we will consider some untraditional statement of the


problem. of values W (z ) , U ( z ) , V (z ) , following (1),

II. STATEMENT OF THE PROBLEM W ( z) U ( z) V ( z) . (2)


Usually the problem reconstruction of signals is considered in
the presence of noise and interpreted as extraction of signal Both component U ( z ) , V (z ) of tandem is assumed function,
Statement of the problem symmetrical relating to some point (symmetry axis) zU , zV
correspondingly. Figure 1-c illustrates the cases of
We have only two unknown smooth and mutually overlapping generalized symmetry of the following form: value of
functions U ( z ) , V (z ) , which are forming exactly function U ( z ) (or V ( z ) ) are predetermined by the value
predetermined smooth source function W ( z ) U ( z ) V ( z ) of function U ( z ) (or V (z ) ) by coefficient of symmetry
(coefficient of reflection) as was described in the assumption
in the section III).
H

68 www.ijeas.org
About the Recognition and Reconstruction Two Unknown Functions From Known Their Tandem

So at A B D E F G H J K =1,
(a) (b) (c) ~
AU U ( z) we will get result V ( z ) V ( z ) W ( z ) ,
max
W ( z) ~ ~ ~
U V ( z) U ( z ) U ( z ) W ( z ) , where U ( z ) and V ( z ) represents
L W AW max
the searched solutions which must be maximally closely to
L q z V
q AV z U ( z ) and V (z ) correspondingly.
q a
(z ) z
q a 2a
zU zV 0
Figure 1
IV. TANDEM RESTORATION PROCEDURE
Now we consider the special computing procedure, which
III. ASSUMPTIONS (HYPOTHESIZES)
represents the base of all further investigations. Figure 2-a
. Distance 2 a between centers z and z of symmetry
A represents smooth function spiral process of tandem
U V
restoration. For the schematic model of pervious concrete
of functions U ( z ) and V (z ) correspondingly is assumed
problem we try to find a single solution for both functions of
1 if yes, A
a am k ( A 0 if no).
tandem.
B . Center point ( zV zU ) / 2 between two centers of
(a)
q U (z)
symmetry U
symmetry is assumed placed in the point k ( B 1 if 10 9

V=W-U
yes, B 0 if no). 6
symmetry U
5

V=W-U

U=W-V
. Symmetry center of function U ( z ) is spaced in the U

V=W-U
2 1

U=W-V
point z k am 1 if
(we assume this hypothesis is true 3 4
symmetry V

U=W-V
0 if no).
yes, 7 8
symmetry V
. Symmetry center of function V (z ) is spaced in the
D V ( z)
point z k am (we assume this hypothesis is true D 1 if 11
symmetry V zU zV
12

yes, D 0 if no). 1 1
2 2 3
(b) 3
4 (c)
E . Points zU and zV are spaced sufficiently far from 5
4
6 6 5
7
7 8 8
borders L . In other words, we assume relation a L . Figure 2
F . the value of function U ( z ) in the start-point z
U ( ) (we will check this hypothesis step by step F 1
The appearance sequence of interpolation nodes (see below)
if yes, F 0 if no). resembles the linear structure of image light sources, caused
. Function ~
G U ( z ) U ( z ) of symmetry of function by one light point-like source spaced in the point z
~ 1 if yes, G
U ( z ) (and U ( z ) ) ( G 0 , if no). between parallel plane mirrors with coordinates z a .
~ Spiral begins (see Fig 2-a) in the point and goes to the left.
H . Function ( z ) ( z ) of symmetry of function V (z )
V V Vertical transitions of spiral line are caused by equation (2),
~ 1 , if yes, H
0 , if no). Simplest cases of
(and V ( z ) ) ( H horizontal transitions of spiral line are caused by concrete
symmetry present V ,U 1 , however one can use more type of symmetry U ( z ) , V ( z ) of searched functions
U ( z ) , V (z ) correspondingly. Table 1 shows the generation of
general form of symmetry U ( z ) U ( z )U ( z ) ,
spiral; sequences of interpolation nodes x { xi } , p { p i } ,
V ( z ) V ( z )V ( z ) (Figure 1-c). Functions U ( z ) , V (z )

q {qi } ( i 1, 2,3,...,13 ), when starting from point z to
(coefficients of symmetry, coefficient of reflection,) must
provide one-to-one correspondence of functions U ( z ) and the left (see Figure 2-a,b).
U ( z ) , V (z ) and V ( z ) , not breaking their smoothness.

, V ( L) W ( z ) max ,
~ ~
J . V ( L) W ( z ) (3)
max
Table I Spiral sequence of interpolation
, U ( L) W ( z )
~ ~
U ( L) W ( z ) at z L . ~
max max x1 p1 q1 W ( x1 ) p1
~
In other words, functions are assumed smoothly going closely x2 2a~ p 2 p1U ( a~ ) q2 W ( x2 ) p2
~
to zero together with their smooth first derivatives (see Figure x3 4a~ q3 q 2 / V (3a~ ) p3 W ( x3 ) q3
1-a). ~
K . Functions W ( z ) ,U ( z ) ,V ( z ) are assumed smooth at x 4 6a~ p 4 p3U (5a~ ) q 4 W ( x4 ) p 4
~
z L . Their characteristic spatial scales are of the same x5 8a~ q5 q 4 / V ( 7 a~ ) p5 W ( x5 ) q5
~
order with L / 2 . x6 10 a~ p6 p5 U (9 a~ ) q 6 W ( x6 ) p 6

69 www.ijeas.org
International Journal of Engineering and Applied Sciences (IJEAS)
ISSN: 2394-3661, Volume-3, Issue-3, March 2016
~
x 7 12 a~ q7 q6 / V (11a~ ) p 7 W ( x7 ) q 7 by variation ( i 1, 2,..., N a ; j 1,2,..., N ; k 1, 2,..., N ) and
~
x8 14 a~ p8 q7 U (13a~ ) q8 W ( x8 ) p8 selection of correct combinations of three numbers
~ ~
~ ai , j , k . Sorting patterns U ( z ) U ( z; ai ; j ; q k ) ,
x9 16 a~ q9 q8 / V (15 a~ ) p 9 W ( x9 ) q 9
~ ~
x10 18 a~
~
p10 p9 U (17 a~ ) q10 W ( x10 ) p10 V ( z ) V ( z; ai ; j ; q k ) , we will search for the tandem
~ ~ ~
q11 q10 / V (19 a~ ) U ( z ), V ( z ) satisfying all above conditions (assumptions,
x11 20 a~ p11 W ( x11 ) q11
hypothesizes) A K .

Table II shows the generation of spiral sequences y { y j } ,

{ j } , { j } ( j 1,2,3,...,13 ) of interpolation nodes, V. MARKERS OF ERRORS
~ ~
when starting from point z to the right (see Figure 2-c). When searching for the single U ( z ) and single V ( z ) we
sort all N a N N 1 patterns (and corresponding all
Table II Spiral sequence of interpolation nodes
y1 1 q1 1 p1 combinations ai , j , k ). Any real combination of
~
y 2 2a~ 2 1 / V ( a~ ) 2 W ( y2 ) 2 key-numbers ai , j , k includes some small or very small
~ errors. We cant measure immediately errors of tandems
y3 4 a~ 3 2 U (3a~ ) 3 W ( y3 ) 3
~ restoration because both functions U ( z ) , V (z ) are assumed
y 4 6a~ 4 3 / V (5a~ ) 4 W ( y4 ) 4
unknown priori.
~
y5 8a~ 5 4 U (7 a~ ) 5 W ( y5 ) 5 However, we have found as empirical fact - an errors in key
~ numbers are always accompanied: (1) by the appearance of
y 6 10 a~ 6 5 / V (9a~ ) 6 W ( y6 ) 6
the fluctuations function on high spatial frequency ~ / a~
~
y 7 12 a~ 7 6 U (11a~ ) 7 W ( y7 ) 7 and its harmonics (the breach of the hypothesis K to
~ smoothness of U ( z ) , V (z ) , see Section III.); (2) by growing
y8 14 a~ 8 7 / V (13a~ ) 8 W ( y8 ) 8
~ of the modules function U ( L ) , V ( L) on edges L of
y9 16 a~ 9 8 U (15 a~ ) 9 W ( y9 ) 9
~ carrier z L (the breach of the hypothesis J , see Section
y10 18 a~ 10 9 / V (17 a~ ) 10 W ( y10 ) 10
~ III). Thereby, we have built the quantitative criterion
y11 20 a~ 11 10 U (19 a~ ) 11 W ( y11 ) 11 ~; ~
[ a~; ] of the choice correct key numbers:

~; ~
[ a~; ] U V , (6)
Table III Renumbering interpolation nodes from
spiral structure into linear structure
spiral
xi ; p( xi ) ; q( xi ) 12 10 8 6 4 2
1
3 5 7 9 11 13

U IUC IUS
2 2

1/ 2
( 2 L)
1 ~ ~
U ( L) U ( L) , (7)


spiral 1/ 2
2 2 2 1 ~ ~
y j ; ( y j ) ; ( y j ) 13 11 9 7 5 3 1 4 6 8 10 12 14 V IVC IVS ( 2 L) V ( L) V ( L) , (8)
linear 2 4 6 8 10 12 14 16 18 20 22 24 26
L ~
zn ; U ( zn ) ; V ( zn ) 1 3 5 7 9 11 13 15 17 19 21 23 25 IUC L U ( z ) cos(z / a~) dz , (9)

L ~
Using Table III and cubic-spline interpolation [5]-[7], we IVC L V ( z ) cos(z / a~)dz , (10)
obtain 26 nodes of interpolation and following continuous
functions L ~
~ IUS L U ( z ) sin(z / a~) dz , (11)
U ( z) interp[cspline(z,u ), z,u,z ] , (4)
~
V ( z) interp[cspline(z, ), z,,z ] . (5) L ~
IVS L V ( z ) sin(z / a~)dz . (12)
Where z { z n } , U {u n } , { n } ( n 1, 2,3,... ,26).

Spatial period of sequence z { z n } is 2 a . Note, that Sorting all values a i , j , k ( i 1, 2,..., N a ;
calculation procedures represented in Tables 1,2,3 guaranteed
j 1,2,..., N ; k 1, 2,..., N ), we search for the minimum of
(at zero noise 0 ) the obvious equations
the calculated value . But value (and its minimum) not
~
U ( z ) V ( z ) W ( z ) in the nodes z z n (at any a~ , ~ , !) , ~ ~
error of restoration of searched tandem U ( z ) , V ( z ) . The
~ ~
which does not any guarantees for equations U ( z ) U ( z ) 0 , minimum of the value [a~; ~; ] , reached by choosing of three
~ ~
V ( z) V ( z) 0 . correct key-numbers a~ a i a , ~ j , k ,
Now our aim is to find a single solution, V (z ) for tandem give us the tandem, most closing to the correct true functions

70 www.ijeas.org
About the Recognition and Reconstruction Two Unknown Functions From Known Their Tandem

5
~ ~
U U , V V 10 W (13) 3 e [a ; ; ] 2 e [a; ; ] e [a; ; ]
max 2
(see below). 2
1
1
1
Sensitivities (U , a~ ) , (V , a~ ) , (U , ~) ,
~) , (V , 0 0 0
~ ~ 0.5 0 0.5 0 .4 0.4 0.4 q 0.4
(U , ) , (V , ) ((14)-(19)) of restored functions a
~ ~ ~ ,
U ( z ) , V ( z ) to the key-numbers errors a a~ a , Figure 3
~
we can use only after tandem reconstruction (post
~ ~
factum, using true model functions U ( z ) , V (z ) and on the Results U ( z ), V ( z ) of restoration of searched functions
~
base of key-numbers hypothesizes a~ , ~ , and sorting of U ( z ), V ( z ) (a) from their tandem W (z ) (b), using exact
many patterns) given by minimums of errors markers values a , , , are represented in Fig. 4.
~; ~

Source function W (z)


[a~;

Sea rched solutions U( z), V(z )


] ).

~
3 3
(a) (b)
~
U

~
~ 2 2
(U , a ) Lim Max U U /a, (14) L (t ) 1
a 0 W

1
z L a( t )
F (t )
0 0
~ 1
~
V 1
(V , a ) Lim Max V V / a , (15)
a 0
z L 50 0 50 z 50 0 50 z
~ t~ 5 at right choices
t three values
U U , V V 10 W
~ max a 0, 0, 0
(U , ) Lim Max U U / , (16)
0
z L Figure 4

~ One can note, that with any errors U ( z ) U ( z ) ,


~
(V , ) Lim Max V V / , (17) ~
0 V ( z ) V ( z ) of restoration, the sum of tandem U ( z ) V ( z ) is
z L
saved equal W (z ) , following the calculation procedure
~ described in Section IV (excluding, of course, the
(U , ) Lim Max U U / , (18) interpolation errors).
0
z L ~ ~
Results U ( z ), V ( z ) of restoration of searched functions
~
~ U ( z ), V ( z ) (a) and errors of restoration U ( z ) U ( z ) ,
(V , ) Lim Max V V / . (19)
~
V ( z ) V ( z ) (b), admitting only one nonzero error a 0 ,
0
z L ( (U , a ) (V , a ) 0.4 , see (14), (15), are represented in
Further we will consider several examples of tandems Fig. 6.
components with different types of symmetry, with different
Errors o f recon structed functio ns

relations of amplitudes, with different velocity of the decline


Sea rched s olutio ns U (z),V (z )

to edge of the carrier. 3


~

(a) ~ (b)
~ 0.4
V V
2 U
~

VI. SYMMETRIC & ANTISYMMETRIC LORENTZS FUNCTIONS X U( z) U ( z)


0.2
1
IN TANDEM 0
X V( z) 0 V ( z)
In this section (and next section V) we consider [2], [3] 0.2 ~
variants of Lorentzs functions. Now consider combination of 1
~ U U
V z 0.4 z
symmetric and antisymmetrical functions
100 50 0 50 100 100 50 0 50 100


z z
U ( z ) 1.8 / 1 0.005( z 2) 2 (20)
error a 0.5 ; correct 0 , 0

V ( z ) 2.2 sin0.1( z 2) / 1 0.005( z 2) 2
(21)
Figure 5

with coefficients of symmetry ~ ~


Results U ( z ), V ( z ) of restoration of searched functions
V ( z ) 1 , U ( z ) 1 (22) ~
U ( z ), V ( z ) (a) and errors of restoration U ( z ) U ( z ) ,
and true key numbers ~
V ( z ) V ( z ) (b), admitting only one nonzero error 0
a 2 , 0 , U ( 0) . (21)
( (U , ) (V , ) 1.25 , see (16), (17)), are represented in
Errors markers [a ; ; ] , [a; ; ] , [a; ; ] are pointing Fig. 6.
~
to the values a~ , ~ , closest with the parameters a , ,
of correct restoration. They are represented in Fig. 3.

71 www.ijeas.org
International Journal of Engineering and Applied Sciences (IJEAS)
ISSN: 2394-3661, Volume-3, Issue-3, March 2016

Searched solutions U(z ),V (z )

Source function W(z)


2 2

Errors of reconstructed fun ctions

~
~
Searched solutions U (z ),V (z)
(a) (b)
V
~ ~
3 1

~
(a) (b) 1 1
~ ~ L (t )
2 U 0.5 V V
0 W (z ) 0
L (t )
1 b ( t) F (t )
0 W
F (t ) 0 g (t ) 1 1
~
1
~
V
0.5 ~
U U
U z z
2 2
1
50 0 50 50 0 50
50 0 z 50 50 0 z 50 ~ 5
at right choices three values
t ~ z
U U , V V 10 W 0, 0, 0
t t m ax a
error 0.2 ; correct a a , 0
Figure 6 Figure 9
~ ~
~ ~ Results U ( z ), V ( z ) of restoration of searched functions
Results U ( z ), V ( z ) of restoration of searched functions ~
~ U ( z ), V ( z ) (a) and errors of restoration U ( z ) U ( z ) ,
U ( z ), V ( z ) (a) and errors of restoration U ( z) U ( z) , ~
~ V ( z ) V ( z ) (b), admitting only one nonzero error a 0
V ( z ) V ( z ) (b), admitting only one nonzero error 0 ,
( (U , a ) (V , a ) 0.5 , see (15), (16)), are represented in
( (U , ) (V , ) 0.5 , see (18), (19)), are represented in Fig. 10.
Fig. 7.

Errors of recon structed functions


Searched solutions U (z ),V (z )
Errors of reconstructed functions
Searched solutions U( z),V (z )

~
~ ~

2
3 1 ~ (a) 0.4 ~ (b)
(a) V U U
~
~ (b)
~ V V 1 0.2
2 U 0.5 L (t ) a( t )
L (t ) b ( t) 0 0
1 F (t ) b ( t)
0
F (t ) 0 g (t ) 0.2
1 ~ ~
0.5 ~ U 0.4 V V
1 ~
V U U
2
1 100 0 z 100 100 0 z 100
50 0 50 z 50 0 50 z t
correct 0 , 0
t
error a 0.8 ;
t
error 0.4 ; correct a a , t 0
Figure 10
Figure 7 ~ ~
Results U ( z ), V ( z ) of restoration of searched functions
~
VII. ANTISYMMETRIC & ANTISYMMETRIC LORENTZS U ( z ), V ( z ) (a) and errors of restoration U ( z ) U ( z ) ,
~
FUNCTIONS IN TANDEM V ( z ) V ( z ) (b), admitting only one nonzero error 0 ,
For the following model functions ( (U , ) (V , ) 4.1 , see (17), (18)), are represented in

V ( z ) 2.2 sin0.1( z 2 / 1 0.005( z 2)2 (24) Fig. 11.
U ( z ) 1.8 sin0.15( z 2) /1 0.005( z 2) 2
(25)
Errors of reconstructed functions

and their coefficients of symmetry


Searched solut ions U (z ),V (z )

U ( z ) 1 V ( z ) 1 (26)
~

2
(a) 0.4 ~ (b)
with coefficients of symmetry ~ V V
V
~

1
a 2 , 0 , U ( 0) . (27) L( t ) a (t )
0.2
Errors markers [a ; ; ] , [a; ; ] , [a; ; ] are pointing to 0 0
~ F( t ) b (t )
the values a~ , ~ , closest with the parameters a , , of 0.2
1 ~ ~
correct restoration Figure 8. U U U
0.4
2
50 0 50 z 50 0 50 z
2
e [a ; ; ] e [a; ; ] 6 e [a; ; ] t t
10
4 error 0.2 ; correct a 0 , 0
1
5 2 Figure 11
0 0
0.5 0 0.5 0.5 0 0.5 0.5 0 0.5 ~
Results U ( z ), V ( z ) of restoration of searched functions
~
a ~
U ( z ), V ( z ) (a) and errors of restoration U ( z ) U ( z ) ,
Figure 8 ~
V ( z ) V ( z ) (b), admitting only one nonzero error 0
~
Results U ( z ), V ( z ) of restoration of searched functions
~ ( (U , ) (V , ) 2.0 , see (19), (20),are represented in
U ( z ), V ( z ) (a) from their tandem W (z ) (b), using exact values Fig. 12.
a , , , are represented in Fig. 10.

72 www.ijeas.org
About the Recognition and Reconstruction Two Unknown Functions From Known Their Tandem

Errors of reconstructed functions


~ ~Searched solutions U( z),V (z ) V ( z ) V ( z ) (b), admitting only one nonzero error a 0 ,
2
(a) 0.4 ~ (b) ( (U , a ) (V , a ) 8.1 see (15), (16)) are represented in
V V
1
0.2 Fig. 15.
L (t ) a (t )
0 0
F (t ) b (t )

Errors of reconstructed functions


Searched solutions U(z ),V (z)
1 ~ 0.2 1

~ ~
~ (a) (b)
U 0.4
U U 1.2 1 1 ~
2 0.5 U U
50 0 50 z 50 0 50 z L( z) ~ ~ a ( z)
t t 0.5 V U 0
b ( z)
error 0.3 ; correct a 0 , 0 F( z)
0.5 ~
Figure 12 0.2 0 1 V V
1
0.2 0 0.2 0.2 0 0.2
z z
VIII. SEPARATION OF TWO VERY CLOSELY SPACED A z B A z B
RADIO-ASTRONOMIC POINT-LIKE SOURCES error a 0 .001 ; correct 0 , 0
For simplicity we consider one-dimensional aperture antenna Figure 15
with characteristic directivity pattern f ( z ) sinc( 35 z ) (all
~ ~
variables are assumed dimensionless). The last means Results U ( z ), V ( z ) of restoration of searched functions
amplitude response of antenna to some source, spaced in ~
U ( z ), V ( z ) (a) and errors of restoration U ( z ) U ( z ) ,
antennas far zone, z is angle coordinate. We spaced two ~
V ( z ) V ( z ) (b), admitting only one nonzero error 0 ,
pointlike sources in the directions with coordinates
z a 0.005 and z a 0.005 . Angle distance 2 a ( (U , ) (V , ) 0.6 see (17), (18)) are represented in
between two sources is less, than the width of directivity Fig. 16.

Errors of reconstructed functions


pattern main petal. Output of antenna [4] are the Searched solutions U (z ),V (z )

powers U ( z) sin35( z a)2 , V ( z) 0.64sin35( z a)2 ,


~ ~
1
(a) ~ (b)
1 ~ V V
created by sources of the antennas output (amplitude of V
U
~ 0.5
source in the direction z a 0.005 is 1, other source has L (z ) a ( z)
0.5 0
F (z ) b ( z)
amplitude 0.8), V 1 , U 1 , 0 . Errors markers
0 0.5 ~
(6)-(12) [a ; ; ] , [a; ; ] , [a; ; ] are pointing to the U U
~ 1
values a~ , ~ , closest with the parameters a , , of 0.2 0.1 0 0.1 0.2 0.2 0.1 0 0.1 0.2

correct restoration Figure 13 and errors a a~ a , ~ , z


error 0. 3 ; correct a 0, 0 z

~
are going to zero if we are sorting patterns at . In Figure 16
the case considered the smallness of angle parameter a
~ ~
means antennas resolution (or superresolution). Results U ( z ), V ( z ) of restoration of searched functions
~
U ( z ), V ( z ) (a) and errors of restoration U ( z ) U ( z ) ,
1 .5 e [ a ; ; ] 0 .2 e [a; ; ] 6
e [a; ; ]
~
V ( z ) V ( z ) (b), admitting only one nonzero error 0 ,
1 4 ( (U , ) (V , ) 0.25 , see (19), (20)) are represented in
0 .1
0 .5 2 Fig. 17.
0 0 0 0.005
0 0 .01 0.02 0.005 0 0.005 0 0.005
Errors of reconstructed functions


Searched solutions U (z ),V (z )

a
~ ~

1
Figure 13 (a) 1 ~ (b)
1
0.5
V V
~ a ( z)
~ ~
L (z ) V ~
Results U ( z ), V ( z ) of restoration of searched (true) F (z )
0.5 U b ( z)
0

functions U ( z ), V ( z ) (a) from their tandem W (z ) (b), using 0.5 ~


0 1
U U
exact key-numbers a , , are represented in Figure 14. 1
0.1 0 0.1 0.2 0.1 0 0.1 0.2
Searched s olutions U (z ),V (z )

z z
Source function W(z)
~ ~

z A z B
(a) 1.5 (b) error 0 .2 ; correct a 0 , 0
1

L (z ) ~ ~ 1 Figure 17
0.5
V U W( t ) W
F (z )
0.5

0 0 IX. TWO GAUSSIAN FUNCTIONS


0.2 0.1 0 0.1 z 0.2 0.2 0.1 0 0.1 z 0.2
In this section we will test the restoration procedure
~ ~z 5 at right choices tthree values
U U , V V 10 (section IV) on the case of two Gaussian functions
W
m ax a a, 0 , 0
U ( z ) 2.2 exp[ 0.005( z 2) 2 ] (32)
Figure 14
V ( z ) 1.8 exp[0.005( z 2) 2 ] (33)
~ ~ with coefficients of symmetry
Results U ( z ), V ( z ) of restoration of searched functions
~
U ( z ), V ( z ) (a) and errors of restoration U ( z ) U ( z ) , U ( z ) 1 , V ( z ) 1 . (34)

73 www.ijeas.org
International Journal of Engineering and Applied Sciences (IJEAS)
ISSN: 2394-3661, Volume-3, Issue-3, March 2016

Errors of reconstructed functions


And true combination of key-numbers

Searched solutions U (z ),V (z )


~ ~
a 2 , 0 , U ( 0) . (35)
Errors markers [a ; ; ] , [a; ; ] , [a; ; ] are pointing to (a) (b)
2 ~ 0.4 ~ 4
~ U U U
the values a~ , ~ , closest with the parameters a , , of ~ 0.2 3
L (t ) a( t )
correct restoration are represented in Figure 18 1 V W( t ) 2
0
F (t ) b ( t)
0.2 ~ 1
0.8
e [a; ; ]
2
e [a; ; ] 0 V V
e [a ; ; ] 10 0.4 z 0

0.4 1 50 0 50 z 50 0 50 50
5
error t 0.2 correct a 0t , 0
0 0 0
1 0 0 .4 0 0.4 0 .4 0 0.4 Figure 21
a 1
Figure 18
~ ~
~ ~ Results U ( z ), V ( z ) of restoration of searched functions
Results U ( z ), V ( z ) of restoration of searched functions ~
U ( z ), V ( z ) (a) and errors of restoration U ( z ) U ( z ) ,
U ( z ), V ( z ) (a) from their tandem W (z ) (b), using exact values ~
V ( z ) V ( z ) (b), admitting only one nonzero error 0
a , , , are represented in Fig. 19.
( (V , ) (U , ) 1.1 ( see (19), (20)), are represented in
Searched solutions U(z ),V (z )

Figure 22.
Source function W(z)
~ ~

(a) 1.5 (b)


1

Errors of reconstructed functions


L (z ) ~ ~ 1 ~(z )
0.5
V U W( t ) W Searched solutions U (z ),V
F (z ) 3 1
0.5 (a) (b)
~
~

U ~
V V
0
0
2 0.5
0.2 0.1 0 0.1 z 0.2 0.2 0.1 0 0.1 z 0.2 L( t ) a( t )
~z 5 at right choicest three values 0
~
U U , V V 10 W F( t ) 1 ~ b( t)
m ax a 0 , 0, 0 V ~
0.5 U U
Figure 19 0
1
50 0 50 z 50 0 50 z
~ ~ errort 0.2 correct 0t , a0
Results U ( z ), V ( z ) of restoration of searched functions
~ Figure 22
U ( z ), V ( z ) (a) and errors of restoration U ( z ) U ( z ) ,
~
V ( z ) V ( z ) (b), admitting only one nonzero error a 0 ,
( (U , a ) (V , a ) 19 , see (15), (16)) are represented in X. IDENTICAL AND DIFFERENT SEARCHED FUNCTIONS.
Fig. 20. ASYMMETRIC SOURCE FUNCTION.
Figure 23-a represents the different error markers
e e (a; ; ) , corresponding to searched Lorentzs functions
Searched solutions U( z),V (z )

Errors of solutions

~ (a)
~

0.4 (b) 4 2 2
2
U ~ U ( z ) 1.2[1 0.005( z 2) ]
, V ( z ) 2.2[1 0.005( z 2) ]
~ 0.2 V V
~

3 with different amplitudes (1.2 and 2.2 continuous line) and


L (t ) a( t )
V W( t )2 U ( z ) 2.2[1 0.005( z 2) 2 ] 2
F (t )
1
b ( t)
0 , V ( z ) 2.2[1 0.005( z 2) ]
0.2 ~ 1 ( U V 1 ) with identical amplitudes (2.2 and 2.2, dotted
U U
0
0.4 0 line). We see, that case of different amplitudes (and
100 0 z 100 100 0 z 100 asymmetric
100 0 W (z100) ) gives us more sharp shape of error
~ ~ ~
markers [a t ; ; ] and more easy finding of key-numbers
t
error a 0.5 correct 0 0
t
, a , , , than the case of identical amplitudes (and symmetric
Figure 20 W (z ) ). More great difference (Figure 23-b) in the behavior of
~
error markers [a~; ~; ] we see in the cases of Gaussian
~ ~ functions with different amplitudes (1.2 and 2.2, continuous
Results U ( z ), V ( z ) of restoration of searched functions 2
~ line) U ( z ) 1.2exp[-(( z 2)/6) 2 ] , V ( z ) 2.2exp[-(( z 2)/6) ] and
U ( z ), V ( z ) (a) and errors of restoration U ( z ) U ( z ) ,
~ the pair of Gaussian functions with identical amplitudes
V ( z ) V ( z ) (b), admitting only one nonzero error 0 , 2 2
U ( z ) 2.2exp[-(( z 2)/6) ] , V ( z ) 2.2exp[-(( z 2)/6) ] (2.2
( (V , ) (U , ) 1.2 , see (17), (18)) are represented in and 2.2, dotted line). It is easy to see, that error marker
Figure 21. ~; ~
[a~; ] is quite non-sensitive on interval 0 a 2 .

74 www.ijeas.org
About the Recognition and Reconstruction Two Unknown Functions From Known Their Tandem

6 ~ ; ; ) (a) 60 ~ ; ; ) (b) In other words, Gaussian function WG ( z ) can be represented


( a ( a
by the pair (39) of Gaussian functions U ( z a ) and U ( z a )
with shift a in the interval (38).
a~ a~
0 2 4 0 4 XI. CONCLUSION
2
Figure 23 Main conclusion: to separate both similar continuous smooth
symmetric function of given their tandem (doing it by single
Now we consider symmetric source function W (z ) as only way) we need only three true key-numbers, known before or
one Gaussian function founded by sorting patterns. Represented numerical results

W ( z ) WG ( z ) 4 exp ( z / 40)
2
(36)
have an empirical nature and seemed sufficiently unexpected.
This is why we used many graphs in this paper.
(Figure 24-a) or only one Lorentzs function

W ( z ) WL ( z ) 4 / 1 ( z / 40 )
(37) 2
EFERENCES
(Figure 24-b). Lets initiate the search (above section IV) for [1] H. Poor, An Introduction to Signal Detection and Estimation. New
York: Springer-Verlag, 1985.
tandem of symmetric ( U V 1 ) functions on the same [2] Springer Handbook of Atomic, Molecular, and Optical Physics
(Editors: Drake, Gordon W. F.) Springer, 2005
carrier z L , observing behavior of the marker of the error [3] Oxford Master Series in Atomic, Optical and Laser Physics (Book 7),
~; ~
[a~; ] . In this case amplitudes of pairs of searched 346 pages, Oxford University Press; 1 edition (February 10, 2005)
English.
symmetric functions can be of only identical amplitudes. In [4] Merrill Skolnik, Radar Handbook, Third Edition McGraw-Hill
the case of Gaussian source function WG ( z ) the calculation Education, 2008 pages: 1328.
[5] I.J. Schoenberg, A. Whitney, Sur la positivite des determinants de
procedure of section IV generates single pair U , V of translations de functions de frequence de Polya avec une application au
smooth symmetric own functions, attached to concrete probleme d`interpolation par les functions spline, Comptes Rendus. -
1949. Vol. 228. - P. 1996-1998.
single own number of the continuum spectrum of own
[6] I. J. Schoenberg, On equidistant cubic spline interpolation Bull.
numbers Amer. Math. Soc. Vol. 77, N. 6 (1971), 1039-1044.
[7] I. J. Schoenberg, On Cubic Spline Interpolation at Equidistant Nodes.
3 a 10 . (38) Mathematics Research Center, University of Wisconsin, 1971 pages :
59.

On the other hand, Lorentzs source function generates only


~
one pair U , V , but error marker [a~; ~; ] does not show
the key-number a sufficiently clear.
Arabadzhi, Vladimir V. senior researcher, Ph.D.

(a) (b)
~ ; ; )
8 ( a 10 ( a~ ; ; )
unceasing
spectrum
of pairs
4
4
> 0 .2 a~
a~
<10
http://www.ipfran.ru/staff/Arabadzhi1.html
0 0 http://www.ipfran.ru/english/staff/Arabadzhi_e.html
5 10 10
Education: Gorky State University, 1978, profession-radiophysics
Figure 24 (01.04.03), Ph.D.1994, topic Investigation methods active
cancellation low frequency waves.
Professional interests:
Graph of marker [a~; ; ] in Figure 24 means, for instance, () searching for solutions of boundary value wave problems, which
that calculation procedure (described in Section IV) gives true admit us to design absorbing coating of the thickness mach smaller
solution U ( z ) of the equation than absorbed wavelength and without information on the direction
and frequency band of incident waves. This aim is achieved by use
of very small and very quick controlled elements of coating. These
U ( z a ) U ( z a ) WG ( z ) (39) solutions are parametric and cant be reduced to the combination of
electric chains with parameters constant in time.
(b) design of floating devises, which convert boat oscillations into a
for each value a from interval (38). Taking into account the
periodical sequence of one polarity stream pulses (wave engine).
condition Professional carrier: NIRFI (1978-1981),, IAP(RAS) (beginning
from 1981). Engineer-senior researcher.
U ( z a) U ( z a) , (41)

(guaranteed by Section IV), we can see, that

2U ( z a ) WG ( z ) or 2U ( z a ) WG ( z ) . (42)

75 www.ijeas.org

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy