Welcome to Scribd, the world's digital library. Read, publish, and share books and documents. See more
Download
Standard view
Full view
of .
Save to My Library
Look up keyword
Like this
3Activity
0 of .
Results for:
No results containing your search query
P. 1
Fractional_order_Signal_Processing

Fractional_order_Signal_Processing

Ratings: (0)|Views: 454|Likes:
Published by sanjaykumarTHAPAR
New era of Signal Processing starts.

All practical real-time phenomenon are of fractional/ non-integer in nature.
New era of Signal Processing starts.

All practical real-time phenomenon are of fractional/ non-integer in nature.

More info:

Published by: sanjaykumarTHAPAR on Nov 28, 2010
Copyright:Attribution Non-commercial

Availability:

Read on Scribd mobile: iPhone, iPad and Android.
download as PDF, TXT or read online from Scribd
See more
See less

10/27/2012

pdf

text

original

 
Analysis of Dirichlet and Generalized ‘‘Hamming’’ window functionsin the fractional Fourier transform domains
Sanjay Kumar
a,
Ã
, Kulbir Singh
a
, Rajiv Saxena
b
a
Department of ECE, Thapar University, Patiala, Punjab, India
b
Department of ECE, Jaypee University of Engineering and Technology, Raghogarh, Guna, Madhya Pradesh, India
a r t i c l e i n f o
 Article history:
Received 30 November 2009Received in revised form23 March 2010Accepted 13 April 2010Available online 6 May 2010
Keywords:
Fractional Fourier transformDirichlet windowHamming windowHanning window
a b s t r a c t
A new mathematical model for obtaining the fractional Fourier transforms of Dirichletand Generalized ‘‘Hamming’’ window functions is presented. The different parametersfor the window functions are also obtained with the help of simulations. The fractionalFourier transformation contains an adjustable parameter with which the main lobewidth and correspondingly, the minimum stop band attenuation of the resultingwindow function can be controlled.
&
2010 Elsevier B.V. All rights reserved.
1. Introduction
Window functions are used in harmonic analysis toreduce the undesirable effects related to the spectralleakage. They impact on many attributes of a harmonicprocessor which include detestability, resolution, dy-namic range, confidence and ease of implementation[1].Several standard windows are also used to optimizethe requirements of a particular application in signalprocessing.Window functions have been successfully used invarious areas of signal processing and communicationssuch as, spectrum estimation, speech processing, digitalfilter design, and in other related fields such as, beam-forming. A complete review of many window functionsand their properties was presented by Harris[2]. Allwindow functions are designed to reduce the side lobes of the spectral output of Fast Fourier transform (FFT)routines. Whilst applying the window function reducesthe side lobe leakage, it causes the main lobe to broadenthus, reducing the resolution. This is a trade-off that has tobe made, one should choose the window function, whichbest suits the application.Recently, the fractional Fourier transform (FrFT) hasbeen developed and utilized by a number of researchers,and being used in almost all applications where Fouriertransforms were used. For example, the FrFT has beenapplied to optimal Wiener filtering and matchedfiltering[3]. Applications of FrFT have also been describedby Bailey and Swartztrauber et al.[4]. Stankovic et al.[5] have used windowed FrFT for the analysis of non-stationary signals. Also, Sharma et al.[6]have carriedout Kaiser and Parzen-cos
6
(
p
) (PC6) window functionanalysis in fractional Fourier domain to show thedependence of window main-lobe width on the order of FrFT and also an alternate methodology is described totune FIR filter transition bandwidth based on FrFT. In thispaper, the FrFT analysis of Dirichlet and Generalized‘‘Hamming’’ window functions has been carried out fordifferent values of the FrFT angle
a
or FrFT order
a
, both of which are related by
a
=
a
p
/2. An attempt has also beenmade to study the variations of the parameters Half MainLobe Width (HMLW), Maximum Side Lobe Level (MSLL)
Contents lists available atScienceDirectjournal homepage:www.elsevier.com/locate/sigpro
Signal Processing
0165-1684/$-see front matter
&
2010 Elsevier B.V. All rights reserved.doi:10.1016/j.sigpro.2010.04.011
Ã
Corresponding author.
E-mail addresses:
 
and Side-lobe fall-off rate (SLFOR) of these windowfunctions with the variation of the parameter
a
. It isfound that with the adjustment of parameter
a
todifferent values, these window functions can attain amaximum main lobe width and SLFOR.The rest of the paper is organized as follows. Section 2gives an overview of the fractional Fourier transform. Anoverview of the window functions and its parameters arediscussed in Section 3. In Sections 4 and 5, the mathema-tical model of Dirichlet and Generalized ‘‘Hamming’’window functions has been derived using fractionalFourier transformation technique. Experimental resultsare presented in Section 6. The conclusive remarks arepresented in Section 7.
2. The fractional Fourier transform
The Fourier transform (FT) is undoubtedly one of themost valuable and frequently used tools in signal proces-sing and analysis[7–9]. Little need be said of theimportance and ubiquity of the ordinary FT in many areasof science and engineering. The fractional Fourier trans-form (FrFT) has been found to have several applications inthe areas of optics and signal processing[10–13]. It alsoleads to the generalization of notion of space (or time) andfrequency domains, which are central concepts of signalprocessing[14–16].The FrFT is a generalization of the conventional FT,which is richer in theory, flexible in application, andimplementation cost is at par with FT. With the advent of FrFT and the related concept, it is seen that the propertiesand applications of the conventional FT are special casesof those of the FrFT. However, in every area where FT andfrequency domain concepts are used, there exists thepotential for generalization and implementation by usingFrFT.The continuous-time fractional Fourier transform of asignal
x
(
) is defined via an integral[10]
 X 
a
ð
u
Þ ¼
1À1
 x
ð
Þ
a
ð
,
u
Þ
dt 
ð
1
Þ
where the transformation kernel
a
(
,
u
) of the FrFT isgiven bywhere
a
indicates the rotation angle of the transformedsignal for the FrFT.The FrFT with
a
=
p
/2 corresponds to the conventionalFourier transform, and the one with
a
=0 corresponds tothe identity operator. Also, two successive FrFT’s withangles
a
and
b
are equivalent to a single FrFT with anangle (
a
+
b
). Hence, the properties of the conventionalFourier transform can be obtained by substituting
a
=
p
/2in the properties of FrFT.
3. Window function and its parameters
Window functions are widely used in digital signalprocessing for the applications in signal analysis andestimation, digital filter design and speech processing[7].In literature many windows have been proposed[8,9]. Thecommon properties of the window functions can besummarized as follows.
They are real, even, nonnegative and time-limited.
Their Fourier transforms have main lobe at the originand side lobes at both sides. These side lobes aredecaying with asymptotic attenuation of 
À
n
as
-
N
where
n
is an integer[1].The parameters of window functions which are gen-erally used for its evaluation are[17–21]:
Maximum Side Lobe Level (MSLL):
This is the peakripple value of the side lobes and it is evaluated fromthe log magnitude plot of transformed window. This isalso known as selectivity amplitude ratio (SAR).
Selectivity (S/2) or Half Main Lobe Width (HMLW):
This is the frequency at which the Main Lobe drops tothe peak ripple value of the side lobes. For conveniencehalf main lobe width (HMLW) or S/2 is computed.
Side-lobe fall-off rate (SLFOR):
This is the asymptoticdecay rate of the side lobe level. This is also calledasymptotic attenuation. Other parameters associatedwith the window functions are Equivalent NoiseBandwidth (ENBW) and Scalloping Losses (SL), whichare out of scope of this paper.Fig. 1presents a plot of the aforementionedparameters for the Dirichlet window function.
a
,
u
ð Þ ¼
 ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1
À
i
cot
ð
a
Þ
2
p
exp
i
2
þ
u
2
2
cot
ð
a
ÞÀ
iut 
cosec
ð
a
Þ
!
if 
a
is not a multiple of 
p
,
d
ð
À
u
Þ
if 
a
is a multiple of 
2
p
,
d
ð
þ
u
Þ
if 
a
þ
p
is a multiple of 
2
p
:
8>>>><>>>>:
ð
2
Þ
Fig. 1.
Log magnitude plot of Dirichlet window to illustrate thedefinition of the parameters HMLW (S/2), MSLL (SAR) and SLFOR.
S. Kumar et al. / Signal Processing 91 (2011) 600–606 
601
 
4. Dirichlet window function
The mathematical analyses of Dirichlet window functionin the fractional Fourier domain is carried in the followingsection. Without loss of generality, let
w
(
) be unity at theorigin, and time-limited to the interval
9
9
r
1/2, i.e.,
w
ð
Þ ¼
1
9
9
r
1
=
20
elsewhere
(
ð
3
Þ
Therefore, the FrFT of 
w
(
) can be written as
a
ð
u
Þ ¼
 ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1
À
i
cot
ð
a
Þ
2
p
exp
iu
2
2cot
ð
a
Þ
Â
1
=
2
À
1
=
2
1 exp
i
2
2cot
ð
a
ÞÀ
iut 
cosec
ð
a
Þ
!
dt 
ð
4
Þ
Now, rewriting the integral of (4),
a
ð
u
Þ ¼
 ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1
À
i
cot
ð
a
Þ
2
p
exp
iu
2
2cot
ð
a
Þ
Â
1
=
2
À
1
=
2
exp
i
2cot
ð
a
Þ ð
À
u
sec
ð
a
ÞÞ
2
Àð
u
sec
ð
a
ÞÞ
2
n o!
dt 
ð
5
Þ
By substituting (
u
sec (
a
))=
R
and changing the limits of the integration in (5),
a
ð
u
Þ ¼
 ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1
À
i
cot
ð
a
Þ
2
p
exp
À
i
2
u
2
tan
ð
a
Þ
Â
ð
1
=
2
ÞÀ
u
sec
ð
a
ÞÀð
1
=
2
ÞÀ
u
sec
ð
a
Þ
exp
i
2cot
ð
a
Þ
R
2
!
dR
ð
6
Þ
Now, solving the integral
ð
1
=
2
ÞÀ
u
sec
ð
a
ÞÀð
1
=
2
ÞÀ
u
sec
ð
a
Þ
exp
i
2
cot
ð
a
Þ
À Á
R
2
 Ã
dR
the following expres-sion results[22],
ð
1
=
2
ÞÀ
u
sec
ð
a
ÞÀð
1
=
2
ÞÀ
u
sec
ð
a
Þ
exp
i
2cot
ð
a
Þ
R
2
!
dR
¼ À
 ffiffiffiffi
p
2
erfi
ð
1
þ
i
Þ
2
 ffiffiffiffiffiffiffiffiffiffiffiffiffi
cot
ð
a
Þ
12
À
u
sec
ð
a
Þ
!&
À
erfi
ð
1
þ
i
Þ
2
 ffiffiffiffiffiffiffiffiffiffiffiffiffi
cot
ð
a
Þ
À
12
À
u
sec
ð
a
Þ
!'
ð
7
Þ
where
erfi
(
 z 
) is an entire analytical function of 
which isdefined in the whole complex
-plane.By rearranging (6) and (7),
a
ð
u
޼ ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1
À
i
cot
ð
a
Þ
8
exp
À
i
2
u
2
tan
ð
a
Þ
Â
erfi
ð
1
þ
i
Þ
2
 ffiffiffiffiffiffiffiffiffiffiffiffiffi
cot
ð
a
Þ
12
À
u
sec
ð
a
Þ
!&
À
erfi
ð
1
þ
i
Þ
2
 ffiffiffiffiffiffiffiffiffiffiffiffiffi
cot
ð
a
Þ
À
12
À
u
sec
ð
a
Þ
!'
ð
8
Þ
Thus, from (8), it can be seen that the FrFT of Dirichletwindow function is directly dependent on the FrFTangle
a
.
5. Generalized ‘‘Hamming’’ window function
The expression for the Generalized ‘‘Hamming’’ win-dow function in time domain is given as[18]
w
ð
Þ ¼
b
þð
1
À
b
Þ
cos
ð
2
p
Þ
9
9
r
1
=
20
elsewhere
(
ð
9
Þ
For
b
=0.50, Hanning window results and for
b
=0.54,Hamming window results.Rewriting (9) in Euler’s form, one gets
w
ð
Þ ¼
b
þð
1
À
b
Þ
e
i
2
p
þ
e
À
i
2
p
2
or,
w
ð
Þ ¼
b
þ
1
À
b
2
ð
e
i
2
p
þ
e
À
i
2
p
Þ ð
10
Þ
Now, because
w
ð
Þ
2
a
a
ð
u
Þ
Therefore,
a
ð
u
Þ ¼
 ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1
À
i
cot
ð
a
Þ
2
p
exp
iu
2
2cot
ð
a
Þ
Â
1
=
2
À
1
=
2
w
ð
Þ
exp
i
2
2cot
ð
a
ÞÀ
iut 
cosec
ð
a
Þ
!
dt 
ð
11
Þ
i
:
e
:
,
a
ð
u
Þ ¼
 ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1
À
i
cot
ð
a
Þ
2
p
exp
iu
2
2cot
ð
a
Þ
Â
1
=
2
À
1
=
2
b
þ
1
À
b
2
ð
e
i
2
p
þ
e
À
i
2
p
Þ
!
Â
exp i
2
2cot
ð
a
ÞÀ
iut 
cosec
ð
a
Þ
!
dt 
ð
12
Þ
Rewriting the integral in (12),
a
ð
u
Þ ¼
b
 ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1
À
i
cot
ð
a
Þ
2
p
exp
iu
2
2cot
ð
a
Þ
Â
1
=
2
À
1
=
2
exp
i
2
2cot
ð
a
ÞÀ
iut 
cosec
ð
a
Þ
!
dt 
þð
1
À
b
Þ
2
 ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1
À
i
cot
ð
a
Þ
2
p
exp
iu
2
2cot
ð
a
Þ
Â
1
=
2
À
1
=
2
exp
i
2
2cot
ð
a
Þþ
it 
ð
2
p
À
u
cosec
ð
a
ÞÞ
!
dt 
þð
1
À
b
Þ
2
 ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1
À
i
cot
ð
a
Þ
2
p
exp
iu
2
2cot
ð
a
Þ
Â
1
=
2
À
1
=
2
exp
i
2
2cot
ð
a
ÞÀ
it 
ð
2
p
þ
u
cosec
ð
a
ÞÞ
!
dt 
ð
13
Þ
Eq. (13) can be rewritten as
a
ð
u
Þ ¼
1
þ
2
þ
3
ð
14
Þ
Here
1
¼
b
 ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1
À
i
cot
ð
a
Þ
2
p
exp
iu
2
2cot
ð
a
Þ
Â
1
=
2
À
1
=
2
exp
i
2
2cot
ð
a
ÞÀ
iut 
cosec
ð
a
Þ
!
dt 
ð
15
Þ
2
¼ð
1
À
b
Þ
2
 ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1
À
i
cot
ð
a
Þ
2
p
exp
iu
2
2cot
ð
a
Þ
S. Kumar et al. / Signal Processing 91 (2011) 600–606 
602

You're Reading a Free Preview

Download
/*********** DO NOT ALTER ANYTHING BELOW THIS LINE ! ************/ var s_code=s.t();if(s_code)document.write(s_code)//-->