Professional Documents
Culture Documents
Abstract
1. Introduction
One of the most frequently used and valuable tools in signal processing is
Fourier analysis. We use it to see and analyze the spectrum of the objects, filter
noises, multiplexing, optics, quantum mechanics, time-frequency analysis, and so on.
Fractional Fourier transform is the fractional order Fourier transform with definition
Eq.(3). It can be seen as the fractional order of Fourier transform, and has many useful
properties to use in applications. The detail goes for part 2.
The linear canonical transform (LCT) is more general than the FrFT. The FrFT
can only do the rotation of the time-frequency distribution. The LCT can do chirp
multiplication, chirp convolution, tilting, dilationetc, but it become more complex
and take more time to calculation.
1
F ( )
2
f (t )e jt dt (1)
1
f (t )
2
F ( )e jt d (2)
2 2
1 j cot j u2 cot j t2 cot jut csc
e e e f (t )dt , if N , N is integer
2
F (u ) f u , if 2N , N is integer (3)
f u , if (2N+1) , N is integer
Since the fractional Fourier transform can be seen as the general case of the
Fourier transform, it has many properties just as like the usual Fourier transform.
Linearity F ( ai fi (t )) ai F ( fi (t ))
i i
Inverse
F 1
F
Commutativity F 1 F 2 F 2 F 1
Associativity
F 1 F 2 F 3 F 1 F 2 F 3
Time shift cos sin 2
f t e
j
F 2
e ju sin F u cos
e f t e
j
F jt 2
e ju cos F u sin
Multiplication d
F t f t j sin u cos F u
du
Differentiation d
F f t cos ju sin F u
du
Parsevals theorem 2
f t dt F u du
2
Now Fig.1,2 show some examples of fractional Fourier transforms. Giving the
1.4 1.4
1.2 1.2
1 1
0.8 0.8
0.6 0.6
F(w)
F(w)
0.4 0.4
0.2 0.2
0 0
-0.2 -0.2
-0.4 -0.4
-5 -4 -3 -2 -1 0 1 2 3 4 5 -5 -4 -3 -2 -1 0 1 2 3 4 5
w w
(a) (b)
1.4 1.4
1.2 1.2
1 1
0.8 0.8
0.6 0.6
F(w)
F(w)
0.4 0.4
0.2 0.2
0 0
-0.2 -0.2
-0.4 -0.4
-5 -4 -3 -2 -1 0 1 2 3 4 5 -5 -4 -3 -2 -1 0 1 2 3 4 5
w w
(c) (d)
1.4 1.4
1.2 1.2
1 1
0.8 0.8
0.6 0.6
F(w)
F(w)
0.4 0.4
0.2 0.2
0 0
-0.2 -0.2
-0.4 -0.4
-5 -4 -3 -2 -1 0 1 2 3 4 5 -5 -4 -3 -2 -1 0 1 2 3 4 5
w w
(e) (f)
Fig. 1 Fractional Fourier transform with the order (a) 0 (b) 0.01 (c) 0.05
(d) 0.2 (e) 0.4 (f) 0.6 The real part is magenta and imaginary part is blue.
4
1.6 1.6
1.4 1.4
1.2 1.2
1 1
0.8 0.8
F(w)
F(w)
0.6 0.6
0.4 0.4
0.2 0.2
0 0
-0.2 -0.2
-0.4 -0.4
-5 -4 -3 -2 -1 0 1 2 3 4 5 -5 -4 -3 -2 -1 0 1 2 3 4 5
w w
(a) (b)
1.6 1.6
1.4 1.4
1.2 1.2
1 1
0.8 0.8
F(w)
F(w)
0.6 0.6
0.4 0.4
0.2 0.2
0 0
-0.2 -0.2
-0.4 -0.4
-5 -4 -3 -2 -1 0 1 2 3 4 5 -5 -4 -3 -2 -1 0 1 2 3 4 5
w w
(c) (d)
1.6 1.6
1.4 1.4
1.2 1.2
1 1
0.8 0.8
F(w)
F(w)
0.6 0.6
0.4 0.4
0.2 0.2
0 0
-0.2 -0.2
-0.4 -0.4
-5 -4 -3 -2 -1 0 1 2 3 4 5 -5 -4 -3 -2 -1 0 1 2 3 4 5
w w
(e) (f)
Fig. 2 Fractional Fourier transform with the order (a) / 4 (b) 7 /16 (c)
/ 2 (d) 3 / 4 (e) 15 /16 (f) The real part is magenta and
imaginary part is blue.
5
The LCT is a more general case than fractional Fourier transform. The fractional
Fourier transform has one parameteralpha, but the LCT has four parameter to adjust
the signal which we are interested in. And the LCT could use some method to reduce
into the fractional Fourier transform. The definition of LCT is:
jd
u2 j
2 a
j t 2
j e2 b e e
ut
b b
f t dt , b 0
Fa ,b ,c ,d (u )
(3.1)
j
cd u 2
d e 2
f d u , b=0
The properties of LCT are alike the fractional Fourier transform, J. J. Ding [2]
has provide a comprehensive discussion and derivation. Since the LCT can be seen as
the general case of the FrFT, the properties are almost alike.
There are many applications of using LCT, such as filter design, optics, and
quantum mechanics. And it can shape the time-frequency distribution of signals,
such as scaling, chirp multiplication, chirp convolution, rotation and tilting. We will
discuss in part 5.
4. Time-Frequency Distribution
Although the Fourier transform can analyze signals, it can not good enough to
know whether the bandwidth is occupied or available. Nowadays, the bandwidth is a
critical issue, since the bandwidth is fixed. If we want to use bandwidth effectively,
we should know the time-frequency distribution. There are many time-frequency
distributions, such as short-time Fourier transform, Gabor transform, Wigner
distribution function (WDF), and etc [3][4]. Here, we take the Gabor transform and
Wigner distribution as examples. The definition of Gabor transform is:
( t )2
G (t , ) e 2
e j x d (4.1)
1
Wg t , g t / 2 g * t / 2 e j d (4.2)
2
6
cos( t ) , 0 t 10
f (t ) cos(3 t ) , 10 t 20
cos(2 t ) , 20 t 30
Fig. 3 show the time-frequency distribution of f(t). We can see that although
WDF has a higher resolution but has too many cross-terms. Another technique is
called Gabor-Wigner transform proposed by J. J. Ding [5]. It is a technique to have
both advantage of the Gabor transform (faster) and the WDF (higher resolution) but
has less cross-term. See Fig. 4.
20 10
8
15
6
10
4
5
frequency
frequency
0 0
-2
-5
-4
-10
-6
-15 -8
-20 -10
0 5 10 15 20 25 30 0 5 10 15 20 25 30
(a) (b)
Fig. 3 The time-frequency distribution of f(t), (a) using Gabor transform, (b) using
WDF.
7
Gabor WDF
10 10
8 8
6 6
4 4
frequency
frequency
2 2
0 0
-2 -2
-4 -4
-6 -6
-8 -8
-10 -10
-10 -8 -6 -4 -2 0 2 4 6 8 10 -10 -8 -6 -4 -2 0 2 4 6 8 10
(a) (b)
Gabor1.5 WDF0.8
10
4
frequency
-2
-4
-6
-8
-10
-10 -8 -6 -4 -2 0 2 4 6 8 10
time (sec)
(c)
Fig. 4 (a) using the Gabor transform (b) using WDF (c) using Gabor-Wigner
transform G (t , ) Wg t ,
1.5 0.8
There are many differences between the Gabor transform and WDF.
Advantage Disadvantage
1. low computation 1. complex value
Gabor 2. the range of the integration is 2. lower resolution
transform limited
3. no cross term
4. linear operation
1. real 1. higher computation
2. higher resolution 2. cross term
WDF 3. if the time/freq limited, time/freq 3. nonlinear operation
8
3
i 3.4 t
Noise : n(t ) 0.7ei 0.032 t
Gs means the Gabor transform of x(t), Gn means the Gabor transform of n(t).
5 5
0
w
0
w
-5 -5
-10 -10
-10 -5 0 5 10 -10 -5 0 5 10
time (sec) time (sec)
(a) (b)
10 original signal
1
0.5
5
0
-10 -8 -6 -4 -2 0 2 4 6 8 10
0
w
after filtering
1.5
1
-5
0.5
-10 -0.5
-10 -5 0 5 10 -10 -8 -6 -4 -2 0 2 4 6 8 10
time (sec)
(c) (d)
Fig. 5 (a) The Gabor transform of x(t) (b) the Gabor transform of x(t)+n(t) (c) after
filtering (d) time domain signals.
After three fractional transforms and three filters, see Fig.5, compare to the
9
u u
a b 2 0
c d
0 0.5
(a) (b)
v v
u u
a b 1 1
c d
a b 1 0 0 1
c d
1 1
(c) (d)
v v
u u
a b 0 1
c d a b cos( / 4) sin( / 4)
1 0 c d
sin( / 4) cos( / 4)
(e) (f)
10
6. Sampling
B csc
We can show that the sampling rate f s must larger than , that is,
Vt sin . Thus, if signals are not time-limited nor bandlimited, we can choose
B
in order to decrease the sampling rate. The same method applies to the LCT, we can
easily derived Vt b.
B
f f
Vf
Vf '
t
t
(a)
(b)
Fig. 7 (a) The time-frequency distribution of original signal, sampling rate must larger
than 2V f (b) after the LCT, sampling rate can reduce to 2V f '
12
7.Fourier Optics
8. Conclusions
References
[1] H.M. Ozaktas, Z. Zalevsky, M. A. Kutay, The fractional Fourier transform with
Applications in Optics and Signal Processing, John Wiley & Sons, 2001.
[2] J. J. Ding, Research of Fractional Fourier Transform and Linear Canonical
Transform, Ph.D. thesis, National Taiwan University, Taipei, Taiwan, R.O.C,
2001.
[3] S. Qian and D. Chen, Joint Time-Frequency Analysis: Methods and Applications,
Prentice Hall, N.J., 1996.
[4] R. L. Allen and D. W. Mills, Signal Analysis: Time, Frequency, Scale, and
Structure, Wiley-Interscience, NJ, 2004.
[5] S. C. Pei, J. J. Ding, Relations between Gabor Transforms and Fractional Fourier
Transforms and Their Applications for Signal Processing, Revised Version: T-SP-
04763-2006.R1
[6] X. G. Xia, On Bandlimited Signals with Fractional Fourier Transform, IEEE
Signal Processing Letters, Vol. 3, No. 3, March 1996