The Discrete Fourier Transform and Its Properties We assume discrete signals in CN , which we index their elements by {x(k)}N

1 . We extend these signals to CZ as N -periodic signals. That is, k=0 x(k) = x(k + N ) for all k Z . We use the standard inner product on CN :
N 1

< u, v >=
i=0

ui v i ,

for all u, v CN ,

so that u
2

< u, u >,

for all u CN .

The DFT of x = (x(0), . . . , x(N 1)) CN is the following signal in CN :

N 1

x(n) =
k=0

x(k)e

2ikn N

The Inverse DFT (IDFT) is obtained as follows 1 x(k) = N

N 1

x(n)e
n=0

2ink N

We recall that the convolution of N -periodic vectors x and y has the form
N 1

(x y)(m) =
k=0

x(m k)y(k) for all m Z ,

and that x y is also N -periodic. We proved the IDFT formula in the following way. We dened the N vectors in CN , {e }N 1 , by =0 e (k) = e
2i k N

, k = 0, . . . , N 1 ,

and expressed the DFT as follows x(n) =< x, en > . 1

We showed that
N 1

e N

N 1 =0

is an orthonormal basis for CN . Therefore,

N 1

x=
n=0

en en 1 < x, > = N N N

n=0

1 < x, en > en = N

N 1

x(n) en ,
n=0

and x(k) = 1 N
N 1 N 1

x(n) en (k) =
n=0 n=0

x(n) e

2ink N

, k = 0, 1, . . . , N 1 .

This proved the IDFT formula. You will also prove in your homework that
N 1

x(k)(k) = y
k=0

1 N

N 1

x(n)y (n) ,
n=0

and thus

1 x N Here are some properties of the DFT x

2

Property Signal DF T Linearity 1 x1 + 2 x2 1 x1 + 2 x2 2ikn0 x(n n0 ) Modulation x(k)e N k Specic modulation x(k)(1) x(n N ) 2 2ink0 Translation x(k k0 ) x(n)e N Translation to center x(k N ) x(n)(1)n 2 Convolution x y(k) x(n)(n) y Multiplication x(k)y(k) x y (n) Correlation x y(k) x(n)y (n) Multiplication by conjugate x(k)(k) y x y(n) Discrete unit impulse (k) 1 Constant impulse 1 (n)

Property Symmetries

Signal x is real

DF T

Re() is even x Im() is odd x x is imaginary Re() is odd x Im() is even x x is real and even x is real and even x is real and odd x is imagninary and odd x is imaginary and even x is imaginary and even x is imaginary and odd x is real and odd