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

A Breif Introduction to the Fractional Fourier Transform

Ratings: (0)|Views: 180|Likes:
Published by Ian Hoover
The Fractional Fourier transform is a conceptually straitforward generalization of the Fourier transform. Given that you can compose the Fourier transform with itself as many times as you want, it is natural to ask the question, which transformation, when composed with itself yields the Fourier transform? Thus we have the Fractional Fourier Transform. This paper derives the specific equation for the fractional transform (in a hopefully straight forward and intelligable way), and investigates the concept of differentiating the fractional transform with respect to the fraction of the Fourier transform we are taking. This paper was the result of my final project for a Signals and Systems class I took at Olin College of Engineering.
The Fractional Fourier transform is a conceptually straitforward generalization of the Fourier transform. Given that you can compose the Fourier transform with itself as many times as you want, it is natural to ask the question, which transformation, when composed with itself yields the Fourier transform? Thus we have the Fractional Fourier Transform. This paper derives the specific equation for the fractional transform (in a hopefully straight forward and intelligable way), and investigates the concept of differentiating the fractional transform with respect to the fraction of the Fourier transform we are taking. This paper was the result of my final project for a Signals and Systems class I took at Olin College of Engineering.

More info:

Categories:Types, School Work
Published by: Ian Hoover on Jun 30, 2013
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

08/30/2013

pdf

text

original

 
Sigsys Final Project Progress Report
Ian HooverMay 5, 2013
My progress has been solely in the increase of my understanding of the fractional Fouriertransform. I will go through what I know.
1 Intro
We showed in class that the Fourier transform preserves the
L
2
norm of functions it actson. This means we can think of the Fourier transform as a map from
L
2
[
−∞
,
] to itself.Indeed, we even talked about the Fourier transform as a basis change (albeit, the baseswere of distributions, so perhaps our space is more accurately that of distributions on
L
2
,which is a superset of functions in
L
2
, but whatever). Now, the upshot of this is that wecan compose the Fourier transform with itself. Thus we can define
n
{
(
t
)
}
=
n
(
u
) by
n
{
(
t
)
}
(
F ◦ F ◦ ··· ◦
)
    
n
{
(
t
)
}
=
n
(
u
)The idea of the fractional Fourier transform is to generalize this notion to non-integer
n
.When we do this, we want, of course, the additive property to hold. That is, we demandthat
α
◦ F 
β
=
α
+
β
If we wanted to calculate the fractional power of a matrix, we would diagonalize it, andthen raise the eigenvalues to said fractional power. The same can be done with the Fouriertransform. Let us suppose we have a countable eigenbasis of 
L
2
given by Ψ
n
(
t
) withassociated eigenvalue
λ
n
. Computing the fractional Fourier transform (henceforth referredto as the FRFT) of one of these basis vectors is trivial. We have that
α
{
Ψ
n
}
=
λ
αn
Ψ
n
We can essentially think of this as our definition of the FRFT, since defining a lineartransformation on the basis of a space defines the transformation on the whole space (with1
 
this definition, it is easy to see that our additive property holds). Now, to calculate theFRFT of an arbitrary function
(
t
), we must first express it in terms of the eigenbasis,
=
1
A
n
Ψ
n
and then apply use linearity to apply the FRFT to each term:
α
{
}
(
u
) =
n
=1
A
n
λ
αn
Ψ
n
(
u
)Now, as is we know, in the discrete case of the Fourier transform, the matrix is diagonal,and thus its eigenvectors are orthogonal. The same is true in the continuous case.
1
Thus,we can compute the basis transformation very simply.
A
n
=
 
−∞
(
t
n
(
t
)
dt
Now we have
α
{
}
(
u
) =
n
=1
A
n
λ
αn
Ψ
n
(
u
) (1)=
n
=1
 
−∞
(
t
n
(
t
)
dt
λ
αn
Ψ
n
(
u
) (2)=
n
=1
 
−∞
(
t
)
λ
αn
Ψ
n
(
t
n
(
u
)
dt
(3)=
 
t
=
−∞
(
t
)
n
=1
λ
αn
Ψ
n
(
t
n
(
u
)
dt
(4)Then, if we define
B
α
(
t,u
) =
n
=1
λ
αn
Ψ
n
(
t
n
(
u
), we have expressed the FRFT as anintegral transformation:
α
{
(
t
)
}
(
u
) =
 
−∞
(
t
)
B
α
(
t,u
)
dt
Thus, finding an elementary expression for the FRFT comes down to finding these functions
B
α
(
t,u
). This can be done since the eigenfunctions Ψ
n
(
t
) are well known. However, thecomputation is not terribly fun. There is a nice way I have found to intuit almost all of what
B
(
t,u
) needs to be though. And for this we need to do a lot of hand-waving, butit is fun hand-waving. Hopefully I will have this on more rigorous footing for the finaldeliverable.
1
The proof of this is really cool, and I will include it at some point in the final thing. Just a bit shorton time at the moment
2
 
2 Waving hands in a great blur
Let us think of the time-frequency plane. We can sort of think of 
δ 
(
t
t
0
) as a verticalline in this space. At time
t
=
t
0
, the thing goes crazy and you have every mode equallyrepresented. Everywhere else it is zero. On the other hand, the function
e
iωt
is of constantfrequency component at all times and is thus represented by a horizontal line. Thus, wecan sort of think of the Fourier transform as a rotation through
π/
2 radians of the time-frequency plane. Now then, the FRFT is just a rotation through arbitrary angle
α
. Ireally should have figures for what comes next, and I will, but bear with me. If we rotatea vertical line of distance
u
from the origin by angle
α
(for the sake of the mental picture,say 0
< α < π/
2). Now, we have a line of decreasing slope. Specifically, the slope of theline is tan(
α
π/
2) =
cot(
α
). Now through basic trig, we find that the
y
-intercept, as itwere, is at
u
csc(
α
). Thus, we kind of have a function, which at time
t
= 0 has frequency
u
csc(
α
), but the frequency decreases by
cot(
α
)
t
. Thus we have
e
ω
(
t
)
t
where we knowthat
ω
(0) =
u
csc(
α
) and that that
dt
=
cot(
α
)
t
. Thus,
ω
(
t
) =
u
csc(
α
)
t
cot(
α
)
t
2
/
2.Well, this is great, we kind of know what happens to a delta function under the FRFT. Itseems like it gets taken to the chirp
e
i
(
u
csc(
α
)
t
cot(
α
)
t
2
/
2
)Which should mean that this isour desired
B
α
(
t,u
). However, we know from our definition of 
B
α
(
t,u
) that it is symmetricin
t
and
u
. Our current chirp function isn’t. To make it so, without messing with ourconditions on
w
(
t
), we simply multiply by
e
i
cot(
α
)
u
2
/
2
and get
e
i
(
u
csc(
α
)
t
cot(
α
)(
t
2
+
u
2
)
/
2
). Itturns out, this is basically what
B
α
(
t,u
) is. However, we want the FRFT to be a unitarytransformation, and for this to be so, we should calculate the normalization factor. Thatis, we want to know
in(
α
{
(
t
)
}|F 
α
{
(
t
)
}
) =
(
|
)where (
·|·
) is the inner product. After a lot of integral manipulation, we arrive at
 
−∞
(
t
)
 
−∞
(
τ 
)
 
−∞
e
i
(
ut
csc(
α
)
cot
(
α
)(
t
2
+
u
2
)
/
2
e
i
(
uτ 
csc(
α
)
cot
(
α
)(
τ 
2
+
u
2
)
/
2
dudτdt
The most inside works out to be 2
π
sin(
α
)
δ 
(
t
τ 
). Thus our normalization factor is
 
12
π
sin(
α
)Which is almost true, except that because of the way inner product is defined, what withmultiplying a number by its complex conjugate and all, what we really have is that
¯
=12
π
sin(
α
)3

Activity (3)

You've already reviewed this. Edit your review.
j. l. hoover added this note
is this 't' supposed to be here?
1 hundred reads

You're Reading a Free Preview

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