Scribd
Upload
36 views

Depi 9

This document summarizes the detection of an analog signal (continuous variant) using a binary transmission system. It describes how the received signal r(t) is modeled as the transmitted signal s(t) plus noise n(t). The optimal receiver theory of Wiener and Holmogroff is discussed, where the goal is to estimate the transmitted signal s(t) from the received signal r(t) by minimizing the mean square error between the estimated signal d^(t) and the desired signal d(t). The solution is derived as an integral equation known as the Wiener equation, which gives the optimal filter transfer function H(ω) that depends on the power spectral densities of the transmitted signal, noise and their cross

Uploaded by

AndreiSandu
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
36 views

Depi 9

This document summarizes the detection of an analog signal (continuous variant) using a binary transmission system. It describes how the received signal r(t) is modeled as the transmitted signal s(t) plus noise n(t). The optimal receiver theory of Wiener and Holmogroff is discussed, where the goal is to estimate the transmitted signal s(t) from the received signal r(t) by minimizing the mean square error between the estimated signal d^(t) and the desired signal d(t). The solution is derived as an integral equation known as the Wiener equation, which gives the optimal filter transfer function H(ω) that depends on the power spectral densities of the transmitted signal, noise and their cross

Uploaded by

AndreiSandu
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 7

Detectia unui semnal (varianta continua)

Sistem de transmisiune binara


S0 semnal transmis = 0 ; t [0, D]
S1 semnal transmis = 1; t [0, D]
D

E = s 2 (t )dt
0

r (t ) = n(t ) ; (S0)
r (t ) = s(t ) + n(t ) ; (S1)
n(t ) - zgomot alb Gaussian
N
n ( ) = 0 ; R
2
1

p (ni ) =

ni2
2 2

e
2
ni = n(KTc ) N 0, 2

r (t ) = lim rk v k (t )
N

{vk (t )}KN =1

= sistem de functii ortogonale

v
0

(1)

K =1

1, k = i
(t )vi (t )dt = Ki =
0, k i

(2)

rk = r (t )v k (t )dt

(3)

v1 (t ) =

s (t )
E
D
1
r1 (t ) =
r (t )s(t )dt
E 0
1

r1 / S 0 =

E
1

r1 / S1 =

(4)
(5)

n(t )s(t )dt =

(6)

[n(t ) + s(t )]s(t )dt =

E+

k>1:
D

rk / S 0 = n(t )v k (t )dt

(8)

rk / S1 = [s (t ) + n(t )]v k (t )dt = n(t )v k (t )dt

rk / S 0 = rk / S1 ; k > 1

(7)

Componentele rk, pentru k>1 nu aduc nici o informatie pentru decizie.


D
1
(9)
statistica suficienta
r1 = l =
r (t )s(t )dt
E 0
p r/ S1
p (r1 , K , rN / S1 ) p (l / S1 )
(10)
r = ~
=
=
= (l )
~
p r/ S 0
p(r1 , K , rN / S 0 ) p (l / S 0 )

( ) (( ))
~

raport de plauzabilitate pentru criteriul Bayes


D
1
l / S0 =
n(t )s(t )dt = (6)
E 0

l / S1 = E +
(7)
Il stim pe si vrem sa determinam p()=?
N

n(t ) n( KT ) sin c K t

(11)

K =1

sin c k t =

T=

sin 2w(t KT )
2w(t KT )

1
; w banda
2w
N

s (t ) s ( jT ) sin c j t

(12)

j =1

N=

l / S0 = =

sin c

1
E

N
N
n
(
KT
)
sin
c
t
k s ( jT ) sin c j t dt (13)

E 0 K =1
j =1

T , k = j
=
(14)
0, k j

n(t )s(t )dt =


0

t sin c j t dt = T Kj

T , k = j

sin
c
t
sin
c
t
dt

K
j
0
0, k j
T N
(13) si (14) => =
n( KT )s( KT )
E k =1
n(KT) Gaussian N(0,T2)
a k n(KT ) Gaussian

Gaussian
=?
var{ } = 2 = ?

n(KT)=0
n aleator (alb, Gaussian)
s determinat
=> = a1 0 + a 2 0 + L = 0

(14)
(15)

Obs n=a1n1 ++aN nN


n1,,nN = variabile aleatoare
n = a1 n1 + L + a N n N
=>

(a)

n 2 = a12 n12 + L + a N2 n N2

(b)

n1 = L = n N = 0 => n = 0

2 =

T2
E

( KT )s 2 ( KT ) (17)

K =1

n ( KT ) = n 2 ( jT ) = 2
2

T 2 N 2
T s ( KT )
=
E K =1
2

E = T s 2 ( KT )

(18)

K =1

1 2 N0
=
2w
2

2 = T 2 =
Obs Pn =

1
2

dom

(19)

( w)dw
n (w)

N0
2

w
-2w

2w
1
Pn = =
2

2w

N ,

p( ) =

N0
N
dw = 0 w 22ww = N 0 w
2
4
w

N0

N0
2

p(l / S 0 ) =
p (l / S1 ) =

N0
2

2
1

N 0
1

N 0

l2
N0

(20)

(l E )2
N0

(21)

p(l / S1 )
(l ) =
e
p(l / S 0 )
ln (l ) =

(l E )2 +l 2
N0

(22)

2l E
E

(23)
N0
N0

>
D1
(25)
ln (l ) ln K =>
<
D0

(24) si (25) => r (t ) s (t )dt


0

Bayes
> N0
< 2

ln K +

D1
E
=>
2
D0

(26)

2
E
(24)
ln (l ) =
r (t ) s (t )dt

N0 0
N0
r(t) semnalul receptionat
s(t) semnalul transmis
E energia lui s(t)
N0/2 densitatea spatiului de putere a zgomotului
K pragul testului Bayes

Estimarea formei semnalului

Ipoteza continua
r(t) = s(t) + n(t)

(1)

d(t) = s(t + )

(2)

[r (t ) s(t + )]2

sa fie minim

r semnal receptionat
s semnal transmis
n zgomot
d semnal dorit

d(t)

Receptor optimal
Teoria lui Wiener si Holmogroff
Se dau:
Bss() = Rss()
Rnn() functia de autocorelatie a zgomotului
Rsr() sau Rsi n()

s(t)

r(t)

canal

^
d(t)

H(w)
h(t)

n(t)

receptor
optimal

zgomot

d(t) semnal ideal la iesirea receptorului optimal


^
d(t) semnal real la iesirea receptorului optimal
Avem 3 variante pt d(t):
1)
2)
3)

d(t) = s(t) filtrare pura


d(t) = s(t+) ; < 0 filtrare cu intarziere
d(t) = s(t+) ; > 0 filtrare cu predictie

(t ) = [d (t ) d (t + )]

= d (t ) d (t )

(3)

h(t) = ?
sau
H(w) = ?
a.i. 2 (t ) sa fie minim!

h(t)
H(w)
^
d(t)

r(t)
)
d (t ) =

r ( )h(t )d =

h( )r (t )d

(4)

filtru real, deci h(t) = 0 ; t < 0

(t ) = [d (t ) d (t + )]
2

= s (t + ) d (t ) = [ s (t + ) h( )r (t )d ] 2

(5)

2 (t ) = s 2 (t + ) 2 h( )r (t ) s(t + )d + h( )h( )r (t )r (t )d .d =

(6)

= Rss (0) 2 h( ) Rs ,r ( + )d + h( )h( ) Rr ,r ( )d .d

OBS. Fie g(t) functia de pondere a unui filtru realizabil oarecare.


Notam: l(t)=h(t)+g(t)
(7)
2
Unde h(t)-functia de pondere l(t)h(t) va determina un mai mare decat aceea a
filtrului de pondere h(t).
+

l = Rs , s (0) 2 h( ) Rs ,r ( + )d 2 g ( ) Rs ,r ( + )d + h( )h( ) Rr ,r ( )d .d +
2

+ 2 h( ) Rs ,r ( )dd +

g ( )h( ) R

r ,r

( )d .d

l 2 = 2 2 [ g ( ) Rs ,r ( + )d h( ) g ( ) Rr ,r ( )d .d ] + 2 g ( ) g ( ) Rr ,r ( )d .d

g ( ) g ( ) R

r ,r

( )d .d = [ g ( )r (t )d ] 2 0

l ;
2

(10)

g ( ) R

(9)

s ,r

( + )d h( ) g ( ) Rr ,r ( )d .d = 0

sa pentru ca g()=0; <0


si Rr,r(-)=Rr,r(-)
+

g ( )[ R

(12)
(13)
+

( + ) h( ) Rr ,r ( )d ]d = 0

(14)

( )d = Rs ,r ( + )

(15)

s ,r

h( ) R

(11)

r ,r

-ecuatia Wiener ; 0

r(t)=s(t)+n(t)
d(t)=s(t+)
Fitrare optimala
(h(t)-solutie a ecuatiei integrale (15))
(15) este conditia necesara si suficienta.
Ecuatia (15) pentru cazul particular al unui filtru nerealizabil:
+

h( ) R

r ,r

( )d = Rs ,r ( + )

< < +

(15)

F{h(t)*Rr,r(t)}=F{Rs,r(t+)}
F{h(t)}F{Rr,r(t)}=F{Rs,r(t+)}
H ( ) r ( ) = s ,r ( )e j
H ( ) =

s ,r ( )e j

r ( )
r(t)=s(t)+n(t)
Rr ,r (t ) = [ s (t ) + n(t )][ s (t + ) + n(t + )]

(16)
(17)
(18)

s (t )n(t + ) = s (t + )n(t ) = 0
Rr ,r ( ) = Rs , s ( ) + Rn ,n ( )

r ( ) = s ( ) + n ( )

(19)

Rs ,r ( ) = s (t )r (t + ) = s (t )[ s (t + ) + n(t + )] = s (t ) s (t + ) + s (t )n(t + ) = Rs , s ( )
Rs ,r ( ) = Rs , s ( )

s ,r ( ) = s ( )
H ( ) =

s ( )
e j
s ( ) + n ( )

(20)
(21) filtrul obtinut nu e
realizabil

You might also like