optimum design of 2-d lowpass fir filters for image processing based on a new algorithm

4
Proceedings of t h e 25th Chinese Control Conference 7-11 August, 2006, Harbin, Heilongjiang Optimum Design o f 2 - D Lowpass F I R Filters F or Image Processing Based o n A N e w Algorithm Yangsheng Chen, Gangfeng Y a n College of Electrical Engineering, Zhejiang University, Hangzhou 310027 E-mail: playmopke 6 1 63.com ygfgLcee.ueducn Abstract: A double sine basis functionneural network f o r t h e design o f 2 - D lowpass filters i s presented. This neural network i s contrived t o have an energy function that coincides with t h e sum-squared error o f t h e approximation problem a t hand a nd b y ensuring that t h e energy i s a monotonic decreasing function, t h e approximation problem c a n b e solved. T h e training theorem i s proposed, a n d design o f t h e 2 D lowpass filters i s improved obviously. I t conquers t h e primary disadvantages of t h e conventional neural networks that t h e convergence speed i s rather low. T h e simulation results indicate that there a r e n o fluctuation both i n t h e passband and stopband, a n d i t attains near ideal filter attenuation characteristics. K ey Words: 2 - D filter, double sine basis, convergence 1 INTRODUCTION A n important research area f o r two-dimensional digital signal processing is t he design o f 2-D FIR filters. Especially, 2-D FI R filters have important applications i n image processing where the phase of 2-D signals often needs t o b e preserved. Therefore, t h e design an d realization problem o f 2 -D F IR filters havebeen receiving a great deal o f attention. Traditional approaches t o t he design of such filters include t h e window method[1], i n which a smoothing window i s applied t o t h e Fourier series coefficients o f t h e ideal frequency response; The frequency transformation method[2], i n which a one-dimensional(1-D) prototype filter response is mapped into a function o f t w o frequency variables using a n appropriate transformation function; The frequency-sampling method[3], i n which desired frequency response values a r e specified a t certain sample points i n t h e frequency domain. T h e window method leads t o closed-form solutions with a relatively insignificant amount o f computation. However, t h e performances o f t h e amplitude a n d frequency response a r e n o t good, a n d this method s hard t o realize when t h e order o f t h e system i s high. O f the three methods described above, t h e frequency-sampling method i s probably the least used an d h a s been t h e least studied, because this design techniques essentially involve a 2 - D interpolation problem a nd a r e consequently n o t trivial. T h e optimization approach, on the other hand, i s computationally demanding b ut c a n provide sub-optimal solutions. T h e filters designed using t he SVD-Based method have good performances when t h e orders o f t h e filters a r e low[4]. Bu t this method must deal with complex computation, s o i t is also hard t o realize when the order o f t h e system i s high. T h e WLS method i s easy t o realize, b ut i t ca n n ot control the verge frequencies o f the filters accurately, a n d t h e curve o f t h e amplitude frequency fluctuates obviously [5]. I n this paper, a double sine basis function neural network f or t h e design o f 2 - D lowpass filters is presented. T h e network is contrived t o have a n energy function that coincides with the sum-squared error o f t he approximation problem a t hand an d by ensuring that t he energy i s a monotonic decreasing function, the approximation problem c a n b e solved. Tw solutions a re obtained. I n t h e first t he FI R band-pass filter o f type-3 is designed o n t h e basis o f a specified amplitude response a n d i n t h e second a filter that h a s specified maximum passband a n d stopband errors i s designed. Regarding t o the convergence, t h e training rate takes effects greatly t o the convergence o f almost a l l kinds o f neural networks. I f the training rate is not valued properly, th e convergence speed will be slow, even t h e neural network will not b e convergent. T h e convergence theorem which ensures t he double sine basis function neural network convergent is proposed, a n d t he theorem is proved i n details i n section III. S o t h e training rate c a n b e well selected according t o this theorem. This design method conquers t he main disadvantages o f t h e conventional neural networks that the convergence speed i s rather slow[6-9]. T h e simulation results attained near idea filters characteristics. 2 DESIGN O F 2-DIMENSION LOW PASS F IR FILTERS 2 . 1 Th e Double Sine Basis Function Neural Networks Model T h e double sine basis functionneural networks i n figure 1 are used t o design a 2- D lowpass FI R filters. IEEE Catalog Number: 06EX1310 This work i s supported b y the 15 1 Excellent person with ability Foundations o f Zhejiang Province. 1110

Upload: ketan-tang

Post on 06-Apr-2018

221 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Optimum Design of 2-D Lowpass FIR Filters for Image Processing Based on a New Algorithm

8/3/2019 Optimum Design of 2-D Lowpass FIR Filters for Image Processing Based on a New Algorithm

http://slidepdf.com/reader/full/optimum-design-of-2-d-lowpass-fir-filters-for-image-processing-based-on-a-new 1/4

P r o c ee d i n g s o f t h e 2 5 t h C h i n e s e C o n t r o l C o n f e r e n c e

7 - 1 1 A u g us t , 2 0 0 6, H a rb i n, H e i l o n g j i a n g

Optimum D e s i g n o f 2 - D L o w p a s s FIR F i l t e r s F o r I m a g e P r o c e s s i n g

B a s e d o n A New A l g o r i t h m

Y a n g s h e n g C h e n , G a n g f e n g Y a n

C o l l e g e o f E l e c t r i c a l E n g i n e e r i n g , Z h e j i a n g U n i v e r s i t y , H a n g z h o u 3 1 0 0 2 7

E - m a i l : p l a y m o p k e 616 3 . c o m y g f g L c e e . u e d u c n

A b s t r a c t : A d o u b l e s i n e b a s i s f u n c t i o n n e u r a l n e t w o r k f o r t h e d e s i g n o f 2 - D l o w p a s s f i l t e r s i s p r e s e n t e d . T h i s n e u r a ln e t w o r k i s c o n t r i v e d t o h a v e a n e n e r g y f u n c t i o n t h a t c o i n c i d e s w i t h t h e s u m - s q u a r e d e r r o r o f t h e a p p r o x i m a t i o n

p r o b l e m a t h a n d a n d b y e n s u r i n g t h a t t h e e n e r g y i s a m o n o t o n i c d e c r e a s i n g f u n c t i o n , t h e a p p r o x i m a t i o n p r o b l e m c a n b e

s o l v e d . T h e t r a i n i n g t h e o r e m i s p r o p o s e d , a n d d e s i g n o f t h e 2 - D l o w p a s s f i l t e r s i s i m p r o v e d o b v i o u s l y . I t c o n q u e r s t h ep r i m a r y d i s a d v a n t a g e s o f t h e c o n v e n t i o n a l n e u r a l n e t w o r k s t h a t t h e c o n v e r g e n c e s pe e d i s r a t h e r l o w . T h e s i m u l a t i o nr e s u l t s i n d i c a t e t h a t t h e r e a r e n o f l u c t u a t i o n b o t h i n t h e p a s s b a n d a n d s t o p b a n d , a n d i t a t t a i n s n e a r i d e a l f i l t e ra t t e n u a t i o n c h a r a c t e r i s t i c s .

Key W o r d s : 2 - D f i l t e r , d o u b l e s i n e b a s i s , c o n v e r g e n c e

1 INTRODUCTION

An i m p o r t a n t r e s e a r c h a r e a f o r t w o - d i m e n s i o n a l d i g i t a ls i g n a l p r o c e s s i n g i s t h e d e s i g n o f 2 - D F I R f i l t e r s .E s p e c i a l l y , 2 - D F I R f i l t e r s h a v e i m p o r t a n t a p p l i c a t i o n s i ni m a g e p r o c e s s i n g w h e r e t h e p h a s e o f 2 - D s i g n a l s o f t e nn e e d s t o b e p r e s e r v e d . T h e r e f o r e , t h e d e s i g n a n d

r e a l i z a t i o n p r o b l e m o f 2 - D F I R f i l t e r s h a v e b e e n r e c e i v i n ga g r e a t d e a l o f a t t e n t i o n . T r a d i t i o n a l a p p r o a c h e s t o t h e

d e s i g n o f s u c h f i l t e r s i n c l u d e t h e w i n d o w m e t h o d [ 1 ] , i nw h i c h a s m o o t h i n g w i n d o w i s a p p l i e d t o t h e F o u r i e rs e r i e s c o e f f i c i e n t s o f t h e i d e a l f r e q u e n c y r e s p o n s e ; T h e

f r e q u e n c y t r a n s f o r m a t i o n m e t h o d [ 2 ] , i n w h i c h ao n e - d i m e n s i o n a l ( 1 - D ) p r o t o t y p e f i l t e r r e s p o n s e i s m a p p e di n t o a f u n c t i o n o f t w o f r e q u e n c y v a r i a b l e s u s i n g a n

a p p r o p r i a t e t r a n s f o r m a t i o n f u n c t i o n ; T h e

f r e q u e n c y - s a m p l i n g m e t h o d [ 3 ] , i n w h i c h d e s i r e df r e q u e n c y r e s p o n s e v a l u e s a r e s p e c i f i e d a t c e r t a i n s a m p l e

p o i n t s i n t h e f r e q u e n c y d o m a i n . T h e w i n d o w m e t h o d

l e a d s t o c l o s e d - f o r m s o l u t i o n s w i t h a r e l a t i v e l yi n s i g n i f i c a n t a m o u n t o f c o m p u t a t i o n . H o w e v e r , t h ep e r f o r m a n c e s o f t h e a m p l i t u d e a n d f r e q u e n c y r e s p o n s e

a r e n o t g o o d , a n d t h i s m e t h o d i s h a r d t o r e a l i z e w h e n t h eo r d e r o f t h e s y s t e m i s h i g h . Of t h e t h r e e m e t h o d s

d e s c r i b e d a b o v e , t h e f r e q u e n c y - s a m p l i n g m e t h o d i sp r o b a b l y t h e l e a s t u s e d a n d h a s b e e n t h e l e a s t s t u d i e d ,b e c a u s e t h i s d e s i g n t e c h n i q u e s e s s e n t i a l l y i n v o l v e a 2 - D

i n t e r p o l a t i o n p r o b l e m a n d a r e c o n s e q u e n t l y n o t t r i v i a l .T h e o p t i m i z a t i o n a p p r o a c h , o n t h e o t h e r h a n d , i sc o m p u t a t i o n a l l y d e m a n d i n g b u t c a n p r o v i d e s u b - o p t i m a ls o l u t i o n s . T h e f i l t e r s d e s i g n e d u s i n g t h e S V D - B a s e dm e t h o d h a v e g o o d p e r f o r m a n c e s w h e n t h e o r d e r s o f t h ef i l t e r s a r e l o w [ 4 ] . B u t t h i s m e t h o d m u s t d e a l w i t h

c o m p l e x c o m p u t a t i o n , s o i t i s a l s o h a r d t o r e a l i z e w h e n

t h e o r d e r o f t h e s y s t e m i s h i g h . T h e WLS m e t h o d i s e a s y

t o r e a l i z e , b u t i t c a n n o t c o n t r o l t h e v e r g e f r e q u e n c i e s o f

t h e f i l t e r s a c c u r a t e l y , a n d t h e c u r v e o f t h e a m p l i t u d e

f r e q u e n c y f l u c t u a t e s o b v i o u s l y [ 5 ] .I n t h i s p a p e r , a d o u b l e s i n e b a s i s f u n c t i o n n e u r a l

n e t w o r k f o r t h e d e s i g n o f 2 - D l o w p a s s f i l t e r s i s p r e s e n t e d .T h e n e t w o r k i s c o n t r i v e d t o h a v e a n e n e r g y f u n c t i o n t h a tc o i n c i d e s w i t h t h e s u m - s q u a r e d e r r o r o f t h ea p p r o x i m a t i o n p r o b l e m a t h a n d a n d b y e n s u r i n g t h a t t h ee n e r g y i s a m o n o t o n i c d e c r e a s i n g f u n c t i o n , t h e

a p p r o x i m a t i o n p r o b l e m c a n b e s o l v e d . Two s o l u t i o n s a r eo b t a i n e d . I n t h e f i r s t t h e FI R b a n d - p a s s f i l t e r o f t y p e - 3 i sd e s i g n e d o n t h e b a s i s o f a s p e c i f i e d a m p l i t u d e r e s p o n s e

a n d i n t h e s e c o n d a f i l t e r t h a t h a s s p e c i f i e d maximump a s s b a n d a n d s t o p b a n d e r r o r s i s d e s i g n e d . R e g a r d i n g t ot h e c o n v e r g e n c e , t h e t r a i n i n g r a t e t a k e s e f f e c t s g r e a t l y t ot h e c o n v e r g e n c e o f a l m o s t a l l k i n d s o f n e u r a l n e t w o r k s . I ft h e t r a i n i n g r a t e i s n o t v a l u e d p r o p e r l y , t h e c o n v e r g e n c e

s p e e d w i l l b e s l o w , e v e n t h e n e u r a l n e t w o r k w i l l n o t b e

c o n v e r g e n t . T h e c o n v e r g e n c e t h e o r e m w h i c h e n s u r e s t h ed o u b l e s i n e b a s i s f u n c t i o n n e u r a l n e t w o r k c o n v e r g e n t i sp r o p o s e d , a n d t h e t h e o r e m i s p r o v e d i n d e t a i l s i n s e c t i o nI I I . S o t h e t r a i n i n g r a t e c a n b e w e l l s e l e c t e d a c c o r d i n g t o

t h i s t h e o r e m . T h i s d e s i g n m e t h o d c o n q u e r s t h e m a i nd i s a d v a n t a g e s o f t h e c o n v e n t i o n a l n e u r a l n e t w o r k s t h a tt h e c o n v e r g e n c e s p e e d i s r a t h e r s l o w [ 6 - 9 ] . T h e

s i m u l a t i o n r e s u l t s a t t a i n e d n e a r i d e a f i l t e r s c h a r a c t e r i s t i c s .

2 DESIGN O F 2-DIMENSION LOW PASS FIR

FILTERS

2 . 1 The D o u b l e S i n e B a s i s F u n c t i o n N e u r a l N e t w o r k s

Model

T h e d o u b l e s i n e b a s i s f u n c t i o n n e u r a l n e t w o r k s i n f i g u r e 1a r e u s e d t o d e s i g n a 2 - D l o w p a s s FI R f i l t e r s .

IEEE C a t a l o g N u m b e r : 0 6 E X 1 3 1 0

T h i s w o r k i s s u p p o r t e d b y t h e 1 5 1 E x c e l l e n t p e r s o n w i t h a b i l i t yF o u n d a t i o n s o f Z h e j i a n g P r o v i n c e .

1 1 1 0

Page 2: Optimum Design of 2-D Lowpass FIR Filters for Image Processing Based on a New Algorithm

8/3/2019 Optimum Design of 2-D Lowpass FIR Filters for Image Processing Based on a New Algorithm

http://slidepdf.com/reader/full/optimum-design-of-2-d-lowpass-fir-filters-for-image-processing-based-on-a-new 2/4

MIM2 2

J = Z Z L D ( w l j , w 2 k ) - G ( w l j I w 2 k ) ]j = l k = 1

w h e r e D ( w l j , W 2 k ) i s t h e d e s i r e d a m p l i t u d e r e s p o n s e

a t t h e f r e q u e n c y ( W 1 J ' W 2 k )

T h e m a t r i x f o r m i s g i v e n b y

G ( w j 1 , W 2 k ) = C l ( W l j ) T B C 2 ( w 2 k )( 2 )

F i g . 1 . T h e d o u b l e s i n e b a s i s f u n c t i o n n e u r a l n e t w o r k m o d e l

T h e n e t w o r k s c o n s i s t o f t h r e e l a y e r s , w h i c h are i n p u t

l a y e r C i , m i d d l e l a y e r C 2 a n d o u t p u t

l a y e r H r e s p e c t i v e l y [ 1 0 , 1 1] . T h e d o u b l e s i n e f u n c t i o n was

s e l e c t e d as t h e a c t i v e f u n c t i o n .

2 . 2 D e s i g n o f 2 - D Lowpass FIR F i l t e r s B a s e d on t h eA m p l i t u d e R e s p o n s e

T h e t r a n s f e r f u n c t i o n o f a 2 - D l o w p a s s FI R f i l t e r s o f

d i m e n s i o n N i X N 2 i s g i v e n b y

N, N2

H ( z I , Z 2 ) = h ( n , n 2 ) Z l flZ2f2

n i = ° : n 2 = 0

T h e f r e q u e n c y response o f t h e f i l t e r i s g i v e n b y

N 1 N

H ( e j w l , e j w ) = , h ( n 1 n 2 ) Z f ( n l w l + n w)

n i = ° n 2 = °

=M(wI, w 2 ) e O(W1 w 2 )

w h e r e M ( W 1 , W 2 ) = | H ( e j w 1 , e j W 2 ) a n d

O ( W 1 , W 2 ) = a r g H(e j W i , e j W 2 ) are the a m p l i t u d e and p h a s e

response o f t h e f i l t e r s , r e s p e c t i v e l y . F o r a l i n e a r - p h a s eFI R f i l t e r , t h e i m p u l s e response must b e s y m m e t r i c a l [ l ] ,i . e , i t s v a l u e s must s a t i s f y t h e c o n d i t i o n

h ( n 1 , n 2 ) = h [ ( N -1- n l i ) , (N2 -1- n2)

F o r a f i l t e r w i t h q u a d r a n t a l s y m m e t r y , t h e f r e q u e n c y

r e s p o n s e is g i v e n b y

H ( e j W l , e 1 W 2 ) = M ( w I , w 2 ) e J i l w l 2 W 2

w h e r e M ( w 1 , W 2 ) = I G ( w 1 , W 2 ) 1

G ( W 1 , W 2 ) = b ( f , h ) s i n ( f W l ) s i n ( h w 2 )f = 1 h = 1

P I =N1!2 p 2 = N 2 / 2

E r r o r a t e a c h p o i n t i s g i v e n b y

E ( j , k ) = D ( W l j , W 2 k ) -G ( w 1 j , W 2 k )

(1)

L e t t h e d e s i r e d a m p l i t u d e response b e d e f i n e d a t

mIXm2 g r i d p o i n t s . T h e s u m - s q u a r e d error over t h e s e

f r e q u e n c y p o i n t s i s g i v e n b y

for i = 1 , 2 , , m l , k = 1 , 2 , . . . M2

B = L

b o , o b o i b o , p ,b 1 0 o b 1 ,1,P

. ... b)

..

6 , n , n h n , 1 .. h n , n A -w h e r e L u ' F 1 W v'1PIF,P2j

i s t h e m a t r i x f o r m e d b y t h e f i l t e r c o e f f i c i e n t s .Now

Abf, , Ui

-I UdJ a E ( j , k ) a G ( j , k )

a b f , h d E ( j , k ) d G ( j , k ) d b f , h

= p u E ( j , k ) C l f ( W l j ) C 2 h ( W 2 k )

S o , we g e t

b f , h = b f , h + A b f , h = b f , h + u E ( j , k ) C l f ( W l j ) C 2 h ( W 2 k )

T h e m a t r i x f o r m i s

B =B + u E ( j , k ) C l ( w 1 j ) C 2 ( W 2 k )

w h e r e f U i s t h e t r a i n i n g r a t e .

3 THE CONVERGENCE THEOREM OF THE

DOUBLE SINE BASIS SINE BASIS

FUNCTION NEURALNETWORK

Theorem 1

A - . -8

I f N 1 N2, t h e d o u b l e

n e u r a l n e t w o r k i s c o n v e r g e n t .

P r o o f

s i n e b a s i s f u n c t i o n

L e t L y a p u n o v f u n c t i o n

V ( j , k ) =- E 2 ( j , k )2

we h a v e t h a t

A V ( j , k ) -E (j , k ) + A E ( j , k ) ] - E 2 ( j , k )

= A E ( j , k ) L E ( j , k ) + - A E ( j , k ) ]

( 3 )We g e t

E ( j + l , k + l ) = E ( j , k ) + A E ( j , k )

1111

C 1 C2 H

Page 3: Optimum Design of 2-D Lowpass FIR Filters for Image Processing Based on a New Algorithm

8/3/2019 Optimum Design of 2-D Lowpass FIR Filters for Image Processing Based on a New Algorithm

http://slidepdf.com/reader/full/optimum-design-of-2-d-lowpass-fir-filters-for-image-processing-based-on-a-new 3/4

E ( j , ) k )

: E ( j , k ) + _ a B

w h e r eA - B = - , E ( j , k )

a E ( j , k )

O b v i o u s l y

a B a B

a-E( k )E ( j E k ) ( 2- y E ( j , k ) a E B , k

S u b s t i t u t i n g e q u a t i o n ( 4 ) i n ( 3 ) , we g e t

A V(2j,) = 2£j ,k ) + A E , j , k ) ] 2 E 2 ( j , k )

= A E ( j , k ) E ( j , k ) + I A E ( j , k )2 F

= - j E j , k) aEj) E ( j k ) -I

E ( j , k ) a E ( j , k ) 2

= E 2 ( j , k ) J E ( j , k ) 2r+ 1 9 2 a E ( j , k ) 2 ]

I n o r d e r t o e n s u r e t h e n e u r a l n e t w o r k c o n v e r g e n t , we

g e t

1 2 a E ( j , k ) O

2 a B

S i n c e / > ° , we h a v e

0 < p < 2

aB

From e q u a t i o n s ( 1 ) a n d ( 2 ) , i t f o l l o w s t h a t

a E ( j , k ) a E ( j , k ) G ( j , k ) = - C I ( W l j ) C 2 ( W 2 k )aB a G ( j , k ) a B

S o , we g e t

3 E ( j , k ) T

2 | C ( W l j ) C TP 1 P 2 2

= Z Z C l f ( w l j ) C 2 h ( w 2 k )f = 1 h = 1

P I p 2 2

= X, s i n ( f w 1 j ) s i n ( h w 2 k )f = 1 h = 1

O b v i o u s l y , w h e n

P 1 P 2 2

Z Z s i n ( f w 1 j ) s i n ( h w 2 N ) r N 2f = 1 h = 1

t h e d o u b l e s i n e b a s i s f u n c t i o n n e u r a l n e t w o r k i sc o n v e r g e n t i n a n y c a s e s .

S i n c e

P I p 2 2NN

1< , l s i n ( f w l j ) s i n ( h W 2 k ) l < Ni 2 2f= 1 h = l

we g e t

2= 8N1N2 N1 * N2

4

H e n c e , w h e n

A V ( j , k ) < O

o < / 1 < 8N i * N2

N i N2

4

we h a v e

w h i c h i m p l i e s t h a t t h e n e u r a l n e t w o r k m u s t b e c o n v e r g e n t .T h u s t h e c o n c l u s i o n o f t h e o r e m 1 i s t r u e .

4 COMPUTER SUBMISSION

F i g u r e 2 s h o w s t h e a m p l i t u d e a n d f r e q u e n c y r e s p o n s e o f

t h e i d e a l2 - D FI R l o w p a s s

f i l t e r .T h e d i me n s i o n o f t h e f i l t e r i s c h o s e n t o b e 3 1 x 3 1 , a n d

t h e n e t w o r k i s s i m u l a t e d u n t i l t h e s u m - s q u a r e d e r r o r- 1 0

d e c r e a s e d b e l o w e . T h e p a s s b a n d a n d s t o p b a n d e d g e sa r e c h o s e n t o b e 0 . 5 3 f f . F i g u r e 3 s h o w s t h e a m p l i t u d ea n d f r e q u e n c y p e r f o r m a n c e o f t h e f i l t e r s d e s i g n e d u s i n gt h i s n e u r a l n e t w o r k w i t h t h e s a m e o r d e r a n d t h e same

p a s s b a n d a n d s t o p b a n d e d g e s .C o m p a r e d w i t h f i g u r e 2 , we c a n s e e e a s i l y t h a t t h ep e r f o r m a n c e s o f t h e f i l t e r d e s i g n e d u s i n g t h i s m e t h o d

a p p r o a c h t h e i d e a l 2 - D FI R l o w p a s s f i l t e r .

0

4 4

F i g . 2 . I d e a l a m p l i t u d e r e s p o n s e

1 1 1 2

Page 4: Optimum Design of 2-D Lowpass FIR Filters for Image Processing Based on a New Algorithm

8/3/2019 Optimum Design of 2-D Lowpass FIR Filters for Image Processing Based on a New Algorithm

http://slidepdf.com/reader/full/optimum-design-of-2-d-lowpass-fir-filters-for-image-processing-based-on-a-new 4/4

REFERENCES

U \ r U

2 2

A 2 I r a da1W;rad4 4

F i g . 3 R e a l a m pl i t u d e r e s p o n s e

5 CONCLUSION

T h i s p a pe r s t u d i e st h e

a m p l i t u d e a n d f r e q u e n c y r e s p o n s e

o f t h e 2 - D l o w p a s s F I R f i l t e r s i n d e t a i l s . A d o u b l e s i n eb a s i s f u n c t i o n n e u r a l n e t w o r k w a s p r e s e n t e d t o d e s i g n t h e2 - D l o w p a s s F I R f i l t e r s . T h e a n a l y s i s p r e s e n t e d , s h o w s

t h a t t h e s u m - s q u a r e d e r r o r i s a m o n o t o n i c d e c r e a s i n gf u n c t i o n w h i c h b e c o m e s z e r o s a t a s t a t i o n a r y p o i n t w i t h

p r o p e r t r a i n i n g r a t e , . H e n c e c o n v e r g e n c e t o a s o l u t i o n i s

a s s u r e d . T h e v a l u e o f t r a i n i n g r a t e e f f e c t s t h e c o n v e r g e n c e

s p e e d o f t h e n e t w o r k , e v e n t h e n e t w o r k s a r e n o t

c o n v e r g e n t w i t h i m p r o p e r , u . T h e c o n v e r g e n c e t h e o r e m

w h i c h e n s u r e s t h i s a l g o r i t h m b a s e d o n d o u b l e s i n e b a s i sf u n c t i o n n e u r a l n e t w o r k c o n v e r g e n t i s a l s o p r e s e n t e d , a n d

t h e t h e o r e m i s p r o v e d i n d e t a i l s i n t h i s p a p e r . T h e n t h e

v a l u e o f u c a n b e s e l e c t e d p r o p e r l y . V a l u i n g t h e u n o tb i g g e r o r s m a l l e r w o u l d b e b e t t e r . T h e e x a m p l e g i v e n

a b o v e h a s s h o w n t h a t h i g h e r d e s i g n a c c u r a c y c a n b e

a c h i e v e d . C o m p a r e d w i t h t h e i d e a l f i l t e r i n f i g u r e 2 , wec a n s e e t h a t t h e f i l t e r d e s i g n e d u s i n g t h i s m e t h o d h a s t h ep e r f o r m a n c e s t h a t a p p r o a c h t h e i d e a l f i l t e r . T h i s m e t h o d

r e d u c e d d e s i g n c o m p l e x i t y g r e a t l y , a n d t h e c o n v e r g e n c e

s p e e d i s f a s t . I t c a n a l s o d e s i g n h i g h o r d e r 2 - D F I R

l o w p a s s f i l t e r s e a s i l y .

[ 1 ] W. Lu a n d A . A n t o n i o u s , T w o - D i m e n s i o n a l D i g i t a l F i l t e r s ,D e k k e r , New Y o r k , 1 9 9 2 .

[ 2 ] C . C . B o u s , " D e s i g n o f N - d i m e n s i o n a l m u l t i p l i e r l e s ss p h e r i c a l l y s y m m e t r i c F I R d i g i t a l f i l t e r s b y

t r a n s f o r m a t i o n , " C i r c u i t s , D e v i c e s a n d S y s t e m s , I E E E

P r o c e e d i n g s , v o l . 1 3 5 , n o . 5 , 2 0 3 - 2 1 0 , 1 9 8 8 .[ 3 ] P . A . S t u b b e r u d , " 2 - D l i n e a r p h a s e f r e q u e n c y s a m p l i n g

f i l t e r s a n d 2 - D l i n e a r p h a s e f r e q u e n c y s a m p l i n g f i l t e r s w i t h

f o u r f o l d s y m m e t r y , " 1 9 9 6 I E E E I n t e r n a t i o n a l S y m p o s i u m

o n C i r c u i t s a n d S y s t e ms , 3 5 6 - 3 5 9 , 1 9 9 6 .[ 4 ] T . B . D e n g a n d E . O k a m o t o , " E f f i c i e n t D e s i g n o f

S V D - B a s e d 2 - D D i g i t a l F i l t e r s U s i n g S p e c i f i c a t i o nS y m m e t r y a n d O r d e r - S e l e c t i n g C r i t e r i o n , " I E E E

T r a n s a c t i o n s a n d S y s t e m s , F u n d m e n t a l t h e o r y a n d

A p p l i c a t i o n , v o l . 5 0 , n o . 2 , 2 0 0 3 .

[ 5 ] K . T a k a o a n d I . S a t o s h i , " D e s i g n Of Arma D i g i t a l F i l t e r sW i t h A r b i t r a r y L o g M a g n i t u d e F u n c t i o n By WL ST e c h n i q u e s , " 1 9 8 8 I E E E I n t e r n a t i o n a l C o n f e r e n c e o n

A c o u s t i c s , S p e e c h a n d S i g n a l P r o c e s s i n g , v o l . 3 ,1 4 4 7 - 1 4 5 0 ,

1 9 8 8 .[ 6 ] I . M e h r a n d T . L . S c u l l e y , " A m u l t i l a y e r N e u r a l N e t w o r k

S t r u c d t u r e f o r A n a l o g F i l t e r i n g , " IEEE T r a n s a c t i o n s o n

C i r c u i t s a n d S y s t e m s - I I : A n a l o g a n d D i g i t a l S i g n a lP r o c e s s i n g , v o l . 4 3 , N o . 8 , 6 1 3 - 6 1 8 , 1 9 9 6 .

[ 7 ] R . Z . M o r a w s k i , A . M i e k i n a a n d A . B a r w i c z ,

" C u r v e - f i t t i n g a l g o r i t h m s v e r s u s n e u r a l n e t w o r k s w h e na p p l i e d f o r e s t i m a t i o n o f w a v e l e n g t h a n d p o w e r i n DWDMs y s t e m s , " I n s t r u m e n t a t i o n a n d M e a s u r e m e n t , IEEE

T r a n s a c t i o n s o n , v o l . 5 4 , n o . 5 , 2 0 2 7 - 2 0 3 2 , 2 0 0 5 .

[ 8 ] T . U e d a , K . T a k a h a s h i , I . S a s a s e , a n d S . M o r i ,

" H o p f i e l d - t y p e n e u r a l n e t w o r k s w i t h f u z zy s e t s t o g a t h e rt h e c o n v e r g e n t s p e e d , " 1 9 9 2 I n t e r n a t i o n a l J o i n t C o n f e r e n c e

o n N e u r a l N e t w o r k s , v o l . 4 , 6 2 4 - 6 2 9 , 1 9 9 2 .[ 9 ] R . P o l i k a r , L . U p d a , S . S . U p d a a n d V . H o n a v a r , " L e a r n + + :

a n i n c r e m e n t a l l e a r n i n g a l g o r i t h m f o r s u p e r v i s e d n e u r a ln e t w o r k s , " I E E E T r a n s a c t i o n s o n S y s t e m s , Man a n d

C y b e r n e t i c s , P a r t C , v o l . 3 1 , n o 4 , 4 9 7 - 5 0 8 , 2 0 0 1 .

[ 1 0 ] 0 . S t a n a n d E . K a m e n , "A l o c a l l i n e a r i z e d l e a s t s q u a r e sa l g o r i t h m f o r t r a i n i n g f e e d f o r w a r d n e u r a l n e t w o r k s , " I E E ET r a n s a c t i o n s o n N e u r a l N e t w o r k s v o l . 1 1 , n o . 2 , 4 8 7 - 4 9 5 ,2 0 0 0 .

[ 1 1 ] M . T . H a g a n , H . B . D e m u t h a n d M . H . B e a l e , N e u r a l

N e t w o r k s D e s i g n , P u b l i s h i n g C o m p a n y o f M e c h a n i s m

I n d u s t r y , 2 0 0 3 .

1 1 1 3