Sei sulla pagina 1di 5

I nt ernat i onal Journal of E mergi ng Trends & Technol ogy i n Comput er Sci ence (I JE TTCS)

Web Site: www.ijettcs.org Email: editor@ijettcs.org, editorijettcs@gmail.com


Volume 1, Issue 4, November December 2012 ISSN 2278-6856


Vol ume 1 , I ssue 4 November - December 2 0 1 2 Page 1 6 1

Abstract: This paper represents a new technique for
designing 2D FIR digital filters with flat magnitude and
sharp transition band with low arithmetic complexity.
Considering wide range usage of filters with these parameters
in communication, signal processing, medicine and etc,
nowadays optimized designs of these filters has been
considered. In addition, filters with these details on cut off
band of high pass, band pass and band stop filters with
arbitrary pass band should be done extend. In order to
flatting the pass band, we used Gaussian approximation in
design FIR filters and concluded that ripple in pass band of
filters those were designed using Gaussian approximation
with attenuation constant coefficients in cut off band, were
less than filter designed with Parks_McClellan method and In
order to sharpen the transition band, filters frequency
response modeled by using of trigonometric functions of
frequency and transfer function calculated in time and
frequency domains. Typically we cannot design a 2D
filter from nonsymmetrical impulse response, then we
have create a 2D low pass FIR filter by using of
chebyshev polynomials.
Keywords: : Two-Dimension FIR filter, Flat Magnitude,
Sharp Transition Band, Nonsymmetrical Impulse
Response, Chebyshev Polynomials.
1. INTRODUCTION
Different ways has been applied for designing of two-
dimensional FIR digital filters. One of them is windows
way and another one is frequency transform. McClellan
has introduced frequency transform for symmetrical two-
dimensional FIR digital filters. By this way one-
dimension FIR digital filters transforms to two-dimension
filters by symmetrical impulse response by using of
transform frequency[1]. Also Karam defined two-
dimension filters with complex coefficients by transform
frequency. In this way, designed filters could have
complex coefficients and also he has generalized those
filters to multi-dimension filters. But none of those could
not be used in nonsymmetrical filters [2]. In this paper we
propose a new method for transforming frequency of one-
dimension FIR digital filters with nonsymmetrical
impulse response to state of two-dimension FIR digital
filters.
2. GAUSSIAN APPROXIMATION METHOD IN
DESIGN OF FIR FILTERS
In designing of a digital low pass FIR filter with
transition band of
c
and stop edge frequency of
s
,
filters frequency response present in ideal form in below
formula:
s
c
H

>
s

=
0
1
) ( (1)
One of the ways of approximation of ideal frequency
response is use of Gaussian functions
) ( ) ( ) (
i i r
i
G C H

- =

(2)
Where is the impulse unit function, * is the
convolution introducer and ) (

i
G is normal (Gaussian)
function[3],[4]
2
2
2
2
1
) (


= e G
i
i
(3)
i
In above formula is of Gaussian distribution.
Collection of Gaussian functions that approximate a FIR
filter presented in figure 1.
parameters
i i i
C , ,
formula,
) (
r
H
In should be chosen to
minimize the tolerance of approximation. For that,
Gappizi and his co-workers suggested [3]-[10]

) ( 735 . 0
c s
= (4)

In other word, they said is a constant coefficient and
presented
i
in below formula
num num i i
i
,..., 2 , 1 . 0 . 1 . 2 ,..., = A = (5)

Where
A
was calculated by
( ) = A 1 ln 2 2
2
(6)
Design a linear phase 2D low pass FIR digital
filter with flat magnitude and sharp transition
band with non symmetrical impulse response

Davood Ghaderi
1
, Niloufar Rajabiyoun
2


1,2
Department of Electrical and Electronic Engineering, Islamic Azad University,
International Jolfa Branch, Jolfa, Iran


I nt ernat i onal Journal of E mergi ng Trends & Technol ogy i n Comput er Sci ence (I JE TTCS)
Web Site: www.ijettcs.org Email: editor@ijettcs.org, editorijettcs@gmail.com
Volume 1, Issue 4, November December 2012 ISSN 2278-6856


Vol ume 1 , I ssue 4 November - December 2 0 1 2 Page 1 6 2

)
2
1
( cos
1
1
0
A k
t
c



+ =

) ( 2
z
s
k

=
2
1
+ =
(

A
=
c
c
num
Max num

p p pm
k H

s s + = 0 cos
2
1 ) (
in above formula o is the maximum permissible ripple in
filters pass band,
( )
r
H
in result is
( )
( )

=

=
num
num i
r
i
e
Max
H
2
2
2
2
1

(7)
Also
(8)


By using of Inverse Fourier Transform, filter impulse
response comes from


(9)


Figure 1 Collection of Gaussian functions that
approximate a FIR filter

Fig. 2. presented a digital FIR filter with N=61 (N is FIR
filters order), 3 . 0 =
c
,
33 . 0 =
s

that designed by
Gaussian approximation and is showing a comparison
with a similar Parks_McClellan method and that showing
the ripple in pass band of filters those designed with
using of Gaussian approximation to attenuation constant
coefficients in cut off band are less than similar filters in
which designed with Parks_McClellan method.

3. FILTER DESINING
The proposed model for the pseudo-magnitude of the
filter transfer function is formulated for equiripple pass
band and sharp transition using trigonometric functions
of frequency[11],[12]. In the proposed model for a linear
phase, equiripple pass band, sharp transition, low pass
FIR filter the various regions of the filter response are
formulated as follow.
In the pass band region, the frequency response is:
) 10 (
Where the frequency variable,
( )
pm
H
is the pseudo-
magnitude of the filter response, o is pass band loss,
p
k









(a) (b)
Figure 2 Comparison Gaussian approximation (a) with
Parks_McClellan method (b)

an integer is a filter parameter in the pass band and
p

is
the end of ripple channel frequency.
Transition region spans part of the pass band
( )
p c

as
well as part of the stop band
( )
s z

where
c
is the
cutoff frequency and
s
e
is the stop band edge frequency.
In the transition region, the frequency response is
(11)
z p t pm
k A H s s = ) ( cos ) (
0

Where
t
k an integer is a filter parameter in the
transition region, A is amplitude parameter and is chosen
greater than 1, is the frequency at which ( )
pm
H
equals A,
z

is the frequency at which


( )
pm
H
is zero in
the stop band region.
In the stop band region, the frequency response :

s s =
z z s
s
pm
k H ) ( sin
2
) (
(12)
Where
s
is the stop band loss,
s
k is the filter parameter
in the stop band region.
From(10)

0 cos =
p p
k
(13)

P
p
L
k

2
=
(14)
Where L is odd, i.e., 1,3,5 to give negative slope due to
roll off.
(15)
) ( cos 0 ) (
0
= =
z t pm
k A H

) 16 (
) ( 2
0

=
z
t
k



) ( sin
2 2
) (
z s
s s
pm
k H

= =

(17)

(18)

From (11), ( ) 1 =
pm
H we obtain

(19)


Also,

(20)


Cut-off frequency
(21)


( )
( )
( )
(

+ =

=
num
i
i
n
n e
Max
n h
1
2
2
1
cos
1
2


) ( cos
1
0

=
p t
k
A
)
1
( cos )
1
(
1
0
A k
t
p

=
I nt ernat i onal Journal of E mergi ng Trends & Technol ogy i n Comput er Sci ence (I JE TTCS)
Web Site: www.ijettcs.org Email: editor@ijettcs.org, editorijettcs@gmail.com
Volume 1, Issue 4, November December 2012 ISSN 2278-6856


Vol ume 1 , I ssue 4 November - December 2 0 1 2 Page 1 6 3

} ] cos cos ) 1 [(
) ( 2
sin
2
) cos
1
(cos 2
) ( 2
.
) ( 2
) cos( sin
)
2
1
(
2 2
2
2 2 2 2
z z s
k k
s
s s
p
p z t
t p
p p p
k k
k k
k
k
kA
k
A
A
k k
k k
A
k k
k k
k
k
k
N
h
s


+ =

+
}

+
=

d H n h
pm d
) (
2
1
) (
Stop band edge frequency

) 22 (


Transition region width
(
(
(
(


=

)
2
1
( cos )
2
( cos
1
1 1
A A k
s
t
c s


(23)
In the stop band,
) 24 (
t
z
z
k
k
2
) 1 4 (
0


+
+ =


Where ,... 2 , 1 , 0 k
z
= choose for 0 k
z
= narrowest
transition band of the low pass filter.
The magnitude response
( ) e
pm
H
is as shown in figure 3.
3.1 Impulse Response Coefficients
Let ( ) 1 0 , s s N n n h , be the impulse response of an N-
point linear phase FIR digital filter. The linear phase
condition implies that the impulse response satisfies the
symmetry condition [7],
(25)

The frequency response for a linear phase FIR filter for
odd N is given by:
(26)




Where the pseudo-magnitude response
( )
pm
H
is

(27)

The impulse response sequence determined by this
frequency response is obtained from









(a)








(b)








(c)
Figure 3 (a) Magnitude response of the proposed model
for low pass filter, (b) Magnitude view of the pass band,
(c) Magnitude view of the stop band

(28)

The impulse response coefficients h(n) for the resultant
filter are obtained by evaluating the integral (28) using
equations (10), (11) and (12). Modeling the pseudo-
magnitude response , ( ) e
pm
H
in the pass band region,
transition region and the stop band region respectively.
According that, we obtain the impulse response
coefficients as

(29)







4. Designing Two-dimension FIR Digital
Filters With Flat Magnitude and Sharp
Transition Band
Transfer function of a one-dimension linear digital FIR
filter comes from bellows formula

=
N
n
n
Z n h Z H
0
) ( ) (
(30)
Totally H(z) a filters transformations function suppose
with a nonsymmetrical impulse response where as every
function could be written as set of one even function and
a odd function, therefore H(z) could be written like bellow

) ( ) ( ) ( Z H Z H Z H
a S
+ =
(31)
Which in it ( ) z H
s
related to transformation function of a
filter with a even symmetrical impulse response and
( ) z H
a
related to transform function of a filter is a odd
symmetrical impulse response.
( ) z H
s
and
( ) z H
a
could
be reached by bellow form1 2 ula:


)) ( ) ( (
2
1
) (
1
+ = Z H Z Z H Z H
N
s
(32)
)) ( ) ( (
2
1
) (
1
= Z H Z Z H Z H
N
a
Filters frequency response comes out from below formula
)
2
( cos
1
1
0
A k
s
t
s



+ =
,.... 2 , 1 , 0 , ) 1 ( ) ( = = n n N h n h
) ( . ) (
)
2
1
(

pm
N
j
j
H e e H

=
)
2
1
(
1
cos )
2
1
( 2 )
2
1
( ) (
N
n
pm
n n
N
h
N
h H
z s
s
s
t p
p
k
k k
A A
k
n
h

cos
2
1
2
)
2
1
(
2


+ + =

I nt ernat i onal Journal of E mergi ng Trends & Technol ogy i n Comput er Sci ence (I JE TTCS)
Web Site: www.ijettcs.org Email: editor@ijettcs.org, editorijettcs@gmail.com
Volume 1, Issue 4, November December 2012 ISSN 2278-6856


Vol ume 1 , I ssue 4 November - December 2 0 1 2 Page 1 6 4

( ) ( ) ( ) ( )

0 0
2
a s
N
j
j
jH H e e H + =

(33)

In which in it
( ) ( )
( ) ( ) ( ) ( ) M N n b H
n a H
M
n
n s
M
n
n s
2 cos sin
cos
0
0
0
= =
=

=
=


(34)
Now we can develop ( ) n cos e according to ( ) cos by
chebicev polynomial
( ) ( ) ( ) cos cos
n
T n = (35)
So we can write frequency response of a favorite filter as
below
( ) ( ) ( )
( ) ( ) ( )] cos cos 1
cos [
1
0
2
0

=
=

+ =
M
n
n n
M
n
n n
jM j
T b
T a e e H


(36)
By above formula clearly you can see we can transform
above-mentioned filters to a two-dimension FIR digital
filter by replacing
cos(w)
with two-dimension function
like ( )
2 1
, F . Every function which comes with
( ) 1 ,
2 1
s F
condition could be used. But as a some
limitation of Karam [2] bellows condition must be applied
with above-mentioned condition













(a)











(b)

Figure 4 designed 2-dimension flat magnitude filter
comparison (b) with conventional filters in cutoff
frequency (a)

( ) ( )
2 1 2 1
, , F F = (37)

With above-mentioned condition as an example we can
use bellows function

) cos( ) cos(
2
1
) cos(
2
1
) cos(
2
1
2
1
) , (
2 1 2 1 2 1
+ + + = F
(38)

5. Simulation Results
In this part we present some of simulation which has been
done in this field.



Figure 5 Two-dimension designed filter in cutoff
frequency 25 . 0



Figure 6 designed filter with conventional ones
comparison in cutoff frequency 25 . 0

5. Conclusion
The comparison between Figure 4(a) and 4(b), we can see
the magnitude of the filter which designed by frequency
transformation method is flatter than pass band of filter
which designed by conventional methods. According to
[12] with a trigonometric method we can design a sharp
transition band filter and according to our method a new
model is introduced to produce a 2D low pass FIR filter
with flat magnitude and sharp transition band using
simple mathematical formulas and linear phase. we
concluded that , ripple in pass band of filters designed
using Gaussian approximation with attenuation constant
coefficients in cut off bands are less than filters designed
with parks-McClellan method. Above mentioned design
has been done by MATLAB software, and eventually our
results were close to the theoretical results including
I nt ernat i onal Journal of E mergi ng Trends & Technol ogy i n Comput er Sci ence (I JE TTCS)
Web Site: www.ijettcs.org Email: editor@ijettcs.org, editorijettcs@gmail.com
Volume 1, Issue 4, November December 2012 ISSN 2278-6856


Vol ume 1 , I ssue 4 November - December 2 0 1 2 Page 1 6 5

ripple maximum 0.005db in pass band and 03 . 0 in
transition band.
REFERENCES
[1] J.H.McClellan, The Design of Two-Dimensional
digital filter by transformations, in prov. 7th Annual
Princeton Conf. on Inform. Sci. and Sys., pp.247-
251, 1973.
[2] L.J. Karam, Design pf Complex
multidimensional FIR filters by Transformations ,
IEEE Proceedings.
[3] T.I. Laakso, V. Valimaki, M.Karajalainen, and
U.K Laian, Splitting the unit delay, IEEE Signal
processing Mag. Vol. 13, pp. 30-35, jan 1998.
[4] I.W. Selesnick and C.S. Burrus, Maximally flat
low-pass FIR filters with reduced delay IEEE Trans.
Circuits Syst. II, vol.45, pp. 53-68 , Jan 1998.
[5] S. Samdi, I. Nishihara and H. iwakura, Universal
maximally flat iow-pass FIR systems , IEEE Transe.
Signal processing, vol. 48, pp 1956-1964, july 2000 .
[6] O. Herrmann, un approximation program in no
recursive digital filter design, IEEE Trans Circuit
theory, vol. CT- 18, pp. 411-413, May 1971.
[7] H. Baher, FIR digital filters with simultaneous
conditions on amplitude and group delay Electron.
Lett., vol. 18, pp. 296-301, Apr. 1982.
[8] R. I. Gopinath, Least squared error FIR filters
with flat amplitude or group delay constraints ,
IEEE Signal Processing, Lett. Nov. 2002 .
[9] R. I. Gopinath, Low-pass Delay Filters with Flat
Magnitude and Group Delay Constraints , IEEE
Trans. Signal Processing, vol. 51, pp. 182-192, Jan.
2003.
[10] J. S. Lim, Two Dimensional Signal and Image
Processing , Englewood Cliffs, NG:Premtice Hall,
1990.
[11] Pardeep Kaur, Optimization of FIR Filters
Design using Genetic Algorithm, International
journal of Emerging Trends& technology in
computer science(IJETTSC), Volume 1, Issue 3, pp.
228-232, September October 2012.
[12] J. Rodrigues, Synthesis of Linear Phase Sharp
Transition FIR Digital Filter, ECTI Transaction on
Computer and Information Technology vol.1, no.2
November, pp. 96-99, 2005

AUTHOR

Davood Ghaderi received the B.S and
M.S. degrees in Electrical Engineering from
Islamic Azad University, Tabriz Branch, Iran,
in 2006 and 2009, respectively. From 3 years
ago he had begun a wide research in field of
ideal filters in FIR & IIR domains and
expanding them to optical devices. His favorite
fields are DSP, filters and image and signal
processing and have several papers and survey projects in these
fields. He had taught in different branches of Islamic Azad
University in Iran, such as Jolfa and Ahar branches from five
years ago.

Niloufar Rajabiyoun received the B.S.
and M.S. degrees in Electrical
Engineering from Engineering Faculty
of Islamic Azad University, Tabriz
Branch, Iran, in 2006 and 2009,
respectively. Her main field in M.S. was
speech signals and Blind Source Separation (BSS)
analysis and has extensive study in these fields with more
than 10 different papers and survey projects. She had
taught in different branches of Islamic Azad University
and private institutes of technology in Iran, from four
years ago.