Академический Документы
Профессиональный Документы
Культура Документы
A Discrete-Time Signal
Processing Framework
Introduction
Review of the foundation of
discrete-time signal processing:
Investigation of essential discrete-
time methods
Briefly touch upon the limitations of
these techniques in the context of
speech processing:
Time-frequency uncertainty principle,
and
Theory of time-varying linear systems
c) Digital sequence
Digital DSP
Converter
obtained after
sampling and
c) quantization x[n]
xn A sin 2 n A sin 2F0n A sin 0n
f0
fs
a) b) c)
1
fs
T
Digital-to-Analog Conversion.
a) Processed digital signal y[n].
b) Continuous signal representation ya(nT).
c) Low-pass filtered continuous signal y(t).
n
u[n] [k ]
k
3
2
1
-3 -2 -1 0 1 2 3 n
Unit Impulse/Sample
1 n 0
n , n
0 n 0
[n]
-3 -2 -1 0 1 2 3 n
1 n 2
3n 6 , n
0 n 2
Unit Step
1 n 0
un , n
0 n 0
u[n]
-3 -2 -1 0 1 2 3 n
Unit Ramp
n n 0
r n , n
0 n 0
r[n]
-3 -2 -1 0 1 2 3 n
Exponential Function
xn un, n Ζ & R
-3 -2 -1 0 1 2 3 n -3 -2 -1 0 1 2 3 n
>1 0<<1
-3 -2 -1 0 1 2 3 n -3 -2 -1 0 1 2 3 n
<-1 -1<<0
4 August 2019 Veton Këpuska 31
Discrete-Time Signals (cont.)
The sinusoidal sequence
x[n] = A cos(ωn+Φ)
e j 0 n N e j 0 n
4 August 2019 Veton Këpuska 35
Periodic and A-periodic Discrete-
time Sinusoids
Example
x[n] = (-1)n, ∀n: periodic with fundamental period of N=2
x[n] = u[n], : a-periodic signal
LCM N , M
NM
GCDN , M
GCD Greatest Common Divisor
4 August 2019 Veton Këpuska 39
Special Periodic Discrete-Time
Signals
The sinusoidal sequence
2k
xn A cosn A cos n , or
N
2k
xn A sin n A sin n
N
ω angular frequency of the sequence.
A is magnitude of the sequence.
Φ is the phase offset.
Note that the discrete-time sinusoidal signal is
periodic in the time variable n with period N
only if N = 2πk/ω ∈ Z integer.
shift
4 August 2019 Veton Këpuska 46
Special Periodic Discrete-Time
Signals
Fact:
2k
xn A cosn , where &A0
N
N
is periodic with fundamenta l period
GCD(k, N)
Example:
3
xn A cos n ,
2 6
3 3 * 2 2 3 k 3
2 2*2 4 N 4
4
fundamenta l period is 4
GCD(3,4)
4 August 2019 Veton Këpuska 47
Example
Find the condition for which the given discrete
time signal is periodic.
A cos0 n A cos0 n 0 N
0 N 2k
e j 0 n N e j 0 n
Example:
δ[n] δ[n-
2]
x[n] y[n]
x[n] y[n]
∏(n;n1,n2)
n1 n2
Duration of pulse:
n2-n1+1 samples
4 August 2019 Veton Këpuska 59
Relationship of Pulse with Step &
Unit Impulse Signals
n; n1 , n2 un n1 un n2 1
Example :
n;2,3 un 2 un 4
n2
n; n1 , n2 n k
k n1
Example :
3
n;2,3 2 ... n
n 3 n k
k 2 k 3 k 2
Note : n; n1 , n1 n n1
4 August 2019 Veton Këpuska 60
Train of Pulses
For N≠0
Tn n n kN
k
N
n2
N
n n N
n N
n2
k 2 k 1 k 0 k 1 k 0
T3(n)
Problem 1
Show that r[n]=nu[n-1]
Solution:
Considering that it has been shown that
r[n]=nu[n], the stated problem states
something strange. More specifically it
needs to be shown that nu[n]=nu[n-1]
∀n.
To demonstrate this equality the
Representation Theorem will be utilized:
4 August 2019 Veton Këpuska 67
Solution of the Problem 1
r n nun n uk n k n n k
k 0 k 0
n n n 1 n 2
n n n n 1 n 2
n k
k 1
Comment:
The fact that nδ[n]=0 ∀n, is a special case of
the sifting property of the impulse function:
x[n]δ[n-n0]=x[n0]δ[n-n0] ∀n,n0
Solution
n n0 ; n1 , n2 u[n n0 n1 ] u[n n0 n2 1]
n n n0 n n n0
Solution:
r[2n]=(2n)u[2n]=(2n)u[n]=2nu[n]=2r[n]
Solution
yn xn n kN xn n kN
n n
xkN n kN
n
xkN when n kN
, k 0,1,2,3,
0 otherwise
4 August 2019 Veton Këpuska 73
Discrete-Time System
A discrete–time system can be
thought of as a transformation T(x) of
an input sequence to an output
sequence:
y[n] = T{x[n]}
C
x: Ζ
n x n
input output
T: S S
x[ n ] y[ n ]T { x[ n ]}
input output
It is a mapping within the set of all complex
valued signals S.
Linearity, and
Time invariance,
Principle of superposition.
Additive Property
n
y 2 [ n] x [k ]
k
2
Then
y3[n] = ay1[n] + by2[n], for all a & b.
n
Proof: y3 [ n ] x [k ]
k
3
y[n] x[k ]h[n k ]
k
y[n] T ( x[n]) T ( x[k ] [n k ])
k
T x[k ] [n k ]
k
y[n] x[k ]T ( [n k ]) x[k ]h[n k ]
k k
n
x[k n ]
k
0
k1 k n0
n n0
y1[n] x[k ] y[n n ]
k1
1 0
h[n-k] = h[-(k-n)]
h[n]
n
Causality:
h[n] = 0 for n<0
Stability:
|a| < 1 the system is stable because the impulse
response is absolutely summable:
hn A a
A
n
n n 0 1 a
h1[n]
x[n] h1[n] h2[n] y[n]
x[n] y[n]
x[n] h2[n] h1[n] y[n]
h2[n]
a yn k b xn k
k 0
k
m 0
k
Accumulator Example
n
y[n] x[k ]
k
n 1
y[n 1] x[k ]
k
x[n] y[n]
One-sample
delay
y[n-1]
y[n] = ay[n-1]+x[n]
y[-1] = c
y[-2] = a-1(y[-1]-x[-1]) = a-1 (c+0) = a-1c
…
y[n] = an+1c, n<0
Overall solution
y[n] = an+1c+Kanu[n], n
X ( ) X ( )
X i ( ) X i ( )
n 2
F x[n n0 ] X e jn0
Fe j 0 n
x[n] X 0
n
Note:
FT of x[n] is complex function with unity magnitude
and a linear phase of slope -n0
In time, energy of the signal is concentrated at n=n0
In frequency, the energy is uniformly distributed over
the interval [-π,π].
X ( ) a e n jn
(ae j ) n
n 0 n 0
1 j
X ( ) j
, if ae a 1
1 ae
4 August 2019 Veton Këpuska 109
Example 2.3
If the sequence is multiplied by the complex
exponential ejω0n:
j 0 n 1
e a u[n]
n
j ( 0 )
, a 1.
1 ae
Using linearity (superposition) property of FT and
the previous relation, the FT transform pair for a
real decaying sinewave can be derived:
1 j
cos( ) (e e j )
2
1 1
2a cos( 0 n)u[n]
n
j ( 0 )
j ( 0 )
, a 1
1 ae 1 ae
4 August 2019 Veton Këpuska 110
Example 2.3
Figure 2.2 (next slide) illustrates the implications of
conjugate symmetry of the FT magnitude and phase of
this real sequence:
The magnitude function is even
The phase function is odd.
Note also that faster the change (decreasing the value of a)
of time sequence the broader the positive and negative
frequency components of the signal around the frequencies
ω0 and –ω0 respectively
ak e
k 0
j k n k
2ak e jk ( k )
k 0
Multiple sinusoidals
N N
k
a
k 0
cos( k n k ) k
(a e j k
k 0
( k ) a k e j k
( k ))
n
n n x[n]
2
n
B( x) ( ) 2 X ( ) d ,
2
X ( ) d
2
n - is the average time of the signal.
- is the average frequency of the signal.
n 2
Consequently:
Only complex sequences, or
Only the positive frequencies of a real sequence are used
in computing bandwidth.
n
Where x[n]r may be absolutely summable
when x[n] is not.
n
x[n]r n
Solution:
X ( z) [n n ]z
n
0
n
z n0
Solution:
za
X ( z) ( [n] a [n 1]) z
n
n 1
1 az
z
ROC z and z ≠ 0;
X(z) has a pole at z = 0, i.e., the transform goes to
infinity.
Also X(z) = 0 for z = a, thus it has a zero at a.
1 az 1
,
ROC : az 1 1 z a
(1 a z ) (1 b z)
r k 1
k
1
k
X ( z ) Az Ni
k 1
N0
k ) (1 d k z)
(1
k 1
c z 1
k 1
Ni
Ak
X ( z) 1
k 1 (1 ck z )
k k
jn
y[n] H ( )e
Then Y [ ] X [ ]H [ ]
w[n] W ( )
And
y[n] x[n]w[n]
Then
1
Y ( )
2
X ()W ( )d
1
Y ( ) X ( ) W ( )
2
Where ⊛ denotes circular convolution.
4 August 2019 Veton Këpuska 145
LTI Systems in the Frequency
Domain
Example 2.9 – Consider a
sequence of a periodic train x[n] [n kP]
k
of unit samples:
If x[n] is the input to an LTI
u[n]
system with impulse h[n] a n
1
With Fourier Transform: H ( ) j
, a 1
1 ae
Y [ ] X [ ]H [ ]
1
2 2
j
Y [ ] ( k )
1ae k P P
2 1 2
Y [ ]
P k 1ae j
( k )
P
2 1
Y [ ]
P k j
2
k
1ae P
4 August 2019 Veton Këpuska 147
LTI Systems in the Frequency
Domain – Example 2.9
or
M
Y z k
z k
H [ z] k 0
N
1 k z k
X [ z]
k 1
(1 ak z ) (1 bk z )
1
H ( z ) Az r k 1 Ni k 1
k )
k 1
(1 c z 1
H ( z) H min ( z ) H max ( z )
“Minimum-phase” and “Maximum-phase” terminology applies to discrete-
time signals as well as to systems and their impulse response.
1
, where : a 1
1 az
n 0 n 0
hn n ,0 n M
hn 0, otherwise
4 August 2019 Veton Këpuska 166
FIR Filters
Because h[n] is bounded over the duration 0≤n≤M, it is
causal and stable. The corresponding rational transfer
function reduces to the form:
Mi M0
X ( z ) Az r (1 ak z 1 ) (1 bk z 1 )
k 1 k 1
With Mi+M0=M and with zeros inside and outside the unit
circle; the ROC is the entire z-plane except at the only possible
poles z=0 or z=∞.
FIR filter can be designed to have perfect linear phase.
If we impose on the impulse response symmetry of the form:
h[n]=h[M-n],
then under the simplifying assumption that M is even
H(ω)=A(ω)e-jω(M/2)
Where A(ω) is purely real, implying that phase distortion will not
occur due to filtering – which is an important property in speech
processing.
k 1
N k 0
N 1 2
j
X (k ) x[n]e
kn
N
,0 k N 1
n 0
2
Discrete Time k
0.8
N Discrete
Fourier 0.6
Fourier
Transform 0.4 Transform
1 N 1 2
X ( )e d
jn
x[n] 0.2
2 x[n] 1 X (k )e N ,0 n N 1
j kn
2 N k 0
0 N N 1 2
j kn
X (k ) x[n]e N ,0 k N 1
X ( ) x[n]e
n
jn
-0.2 n 0
-0.4
-0.6
-0.8
-1
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1
N 1 N 1
1
x[n] X (k )
2 2
n 0 N k 0
n 0
2
j kn0
X (k ) e N
Sampling Theorem:
Suppose that xa(t) is:
sampled at a rate of Fs=1/T samples per second.
a band-limited signal: its continuous-time Fourier transform Xa(Ω) is
such that Xa(Ω)=0 for |Ω|≥ ΩN (ΩN = 2FN)
Then xa(t) can be uniquely determined from its uniformly spaced
samples x[n] = xa(nT) if the sampling frequency Fs is greater then
twice the largest frequency of the signal:
Fs > 2FN
sin t nT / T
xa (t ) x (nT )
n
a
t nT / T
H H a ,
1
T T
where ha(t), is assumed to be bandlimited and the
sampling rate satisfies the Nyquist criterion.