Scribd
Upload
32 views

CIS 580 Spring 2013 - Lecture 2: January 16, 2013

1) The Fourier transform represents a function in the time domain as a function in the frequency domain. The Fourier transform of a function decomposes it into its constituent frequencies. 2) Even functions map to the real part of the Fourier transform, while odd functions map to the imaginary part. 3) Convolution in the time domain corresponds to multiplication in the frequency domain - the Fourier transform of the convolution of two functions is equal to the product of their individual Fourier transforms.

Uploaded by

Jonathan Balloch
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
32 views

CIS 580 Spring 2013 - Lecture 2: January 16, 2013

1) The Fourier transform represents a function in the time domain as a function in the frequency domain. The Fourier transform of a function decomposes it into its constituent frequencies. 2) Even functions map to the real part of the Fourier transform, while odd functions map to the imaginary part. 3) Convolution in the time domain corresponds to multiplication in the frequency domain - the Fourier transform of the convolution of two functions is equal to the product of their individual Fourier transforms.

Uploaded by

Jonathan Balloch
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 3

CIS 580 Spring 2013 - Lecture 2

January 16, 2013 Notes and figures by Matthieu Lecce.


Updated by Nicu Stiurca
Last lecture’s main result: linear shift-invariant (LSI) systems can be repre-
sented as a convolution.

The Fourier Transform


Quick reminder on complex numbers:
Definition of the Fourier Transform: • a + jb ∈ C, j2 = −1

f (t ) F (ω) = F { f (t )} • e jωt = cos(ωt ) + j sin(ωt ).

F :R → C
Z ∞
F (ω) = f (t )e− jωt dt,
−∞
where ω denotes the frequency. This definition is sometimes called non-
unitary Fourier transform, with angular frequency (ω is referred to as angular
frequency, and s such that ω = 2πs is the ordinary frequency).
Inverse Fourier transform:
Z ∞
1
f (t ) = F (ω)e jωt dω
2π −∞
Domains:
• f (t ) defined in time (or space for x, y) domain

• F (ω) defined in frequency (or spatial frequency) domain


The Fourier transform can be defined as a function of s, the frequency,
where ω = 2πs, in which case the definitions can be rewritten as follows:
Z ∞
F ( s) = f (t )e− j2πst dt
−∞
Z ∞
1
f (t ) = F ( s)e j2πst ds
2π −∞

Function symmetry and Fourier


Definitions:
• A function fe is said even when fe (−t ) = fe (t ) 4 Example of even function
1
Example of odd function
0.5

• A function fo is said odd when fo (−t ) = − fo (t )


2
0
0
−0.5
−2

Any function f (t ) can be decomposed into an odd and an even part: −5 0 5


−1
−5 0 5

f ( t ) = fe ( t ) + fo ( t ) Figure 1: Even and odd functions.


1
where fe (t ) = ( f (t ) + f (−t )) is even
2
1
fo (t ) = ( f (t ) − f (−t )) is odd
2
cis 580 spring 2013 - lecture 2 2

If we apply the Fourier transform to this decomposition, we obtain the


following:
Z ∞
( fe (t) + fo (t))(cos ωt − jsinωt)dt
Z−∞
∞ Z ∞
= fe (t ) cos(ωt )dt − j ( fo sin ωt)dt
−∞ −∞
R∞
The even part maps to the real part of the Fourier transform and the odd part Observe that g (t )go (t )dt
−∞ e
=0
to the imaginary part (and vice versa).

Theorems
Shift theorem
f (t − t0 )  F (ω)e− jωt0
T
Example: for f (t ) even, f (t − T2 )  F (ω)e− jω 2 , where F (ω) is real.

Modulation theorem
f (t )e jω0 t  F (ω − ω0 )
Multiplying by a complex exponential causes a shift in the frequency domain.

Similarity theorem
1 ω
f (at )  F( )
|a| a

Convolution

This theorem is extensively used in image processing:


R∞
f (t ) → h(t ) → g(t ) = −∞ g(t0 )h(t − t0 )dt0 = f (t ) ∗ h(t )
˜ ˜ ˜
F (ω) H (ω) G (ω) ?
What happens in the Fourier domain?
Z ∞Z ∞
G (ω) = f (t0 )h(t − t0 )dt0 e− jωt dt
−∞ −∞
Z ∞ (Z ∞ )
= f (t 0 ) h(t − t0 )e− jωt dt dt0
t0 =−∞ t =−∞
Z ∞
0
= f (t ) H (ω)e− jωt dt0 (shift theorem)
0
t0 =−∞
= H (ω) F (ω)

H (ω) is call the transition function (as opposed to impulse response).

Inverse convolution
1
f (t )h(t )  F (ω) ∗ H (ω)
2π
cis 580 spring 2013 - lecture 2 3

Figure 2: Low pass filtering by taking the


product H (ω) F (ω).
H(ω) ideal low−pass filter

−2pi −3pi/2 −pi −pi/2 0 pi/2 pi 3pi/2 2pi

F(ω) spectrum of some signal

−2pi −3pi/2 −pi −pi/2 0 pi/2 pi 3pi/2 2pi

Filtered spectrum: F(ω)H(ω)

−2pi −3pi/2 −pi −pi/2 0 pi/2 pi 3pi/2 2pi

You might also like