Professional Documents
Culture Documents
Linear Prediction (LPC)
Linear Prediction (LPC)
Linear Prediction (LPC)
Lecture 6:
Linear Prediction (LPC)
1.
2.
3.
4.
http://www.ee.columbia.edu/~dpwe/e4896/
2013-02-25 - 1 /16
1. Resonance
Resonance is ubiquitous
in physical systems
e.g. plucked/struck
string, drum head
room coloration
vocal tract
A
B
C
D
Poles
easily implemented
in LTI filters
Position
Time
Resonances
0
Velocity
2013-02-25 - 2 /16
imag
Simple resonance =
Second-order IIR
Band Pass Filter
1
0.5
y[n]
z-1
2rcosc
z-1
-r2
r
c
x[n]
-0.5
-1
z-plane
|H(ejw)|
-1
-0.5
0.5
1 real
Q= c
B
B 2(1-r)
0.1
0.2
0.3
impulse_bpf.pd
2013-02-25 - 3 /16
Resonances in Speech
freq / Hz
resonances = formants
4000
3000
2000
1000
h z
0.5
has a
tcl
watch
1.5
time
I z/ s I
thin
as a
ay
m
dime
2013-02-25 - 4 /16
Source-Filter Model
Separation of:
Pitch
Glottal pulse
train
Voiced/
unvoiced
Vocal tract
resonances
Radiation
characteristic
Speech
Frication
noise
t
Source
Filter
Advantages
Good match to real signals
Salient pieces
2013-02-25 - 5 /16
s[n] =
ak s[n
k] + e[n]
k=1
Transfer function
S(z)
=
E(z)
1
1
p
k=1
ak z
1
=
A(z)
2013-02-25 - 6 /16
s[n] =
ak s[n
k] + e[n]
k=1
Pulse/noise
excitation
e[n]
H(z)
Vocal tract
H(z) = 1/A(z)
s[n]
|H(ej)|
z-plane
2013-02-25 - 7 /16
resonances in
a spectral slice:
energy / dB
-100
1000
2000
s[n]
n
freq / Hz
:
2
e2 [n] =
3000
ak s[n
k]
k=1
k]
2013-02-25 - 8 /16
LPC Illustration
0.1
0
windowed original
-0.1
-0.2
LPC residual
-0.3
dB
50
100
150
200
250
300
350
400
time / samp
original spectrum
LPC spectrum
-20
-40
-60
residual spectrum
0
1000
2000
3000
4000
5000
6000
7000
freq / Hz
Actual
poles:
z-plane
E4896_L06_lpc_diary
2013-02-25 - 9 /16
Short-Time LP Analysis
freq / kHz
20
Imaginary Part
10
5
12
-0.5
-1
freq / kHz
15
0.5
-5
-10
-1
0
Real Part
-15
0.2
0.4
0.6
0.8
6
4
2
0
0.5
1.5
2.5
time / s
2013-02-25 - 10/16
3. LP Representations
S(z)
1
=
E(z)
A(z)
E(z) = S(z)A(z)
is unsmoothed S(z)
Signal whitening
2013-02-25 - 11/16
Alternative Forms
Many formulations for pth order all-pole IIR:
predictor coefficients {ak } / polynomial A(z)
roots {
= ri ej i } of A(z)
mathematical convenience
numerical stability
statistical properties (e.g. for coding)
opportunities for modification
2013-02-25 - 12/16
4. LPC Synthesis
Decoder
|1/A(ej!)|
Input
s[n]
Represent
& encode
LPC
analysis
Excitation
generator
Represent
& encode
Residual
e[n]
^
e[n]
All-pole
filter
H(z) =
Output
^
s[n]
1
1 - "aiz-i
16 ms frame boundaries
100
50
0
-50
1.3
1.35
1.4
1.45
1.5
1.55
1.6
1.65
1.7
1.75
time / s
2013-02-25 - 13/16
LPC Warping
Replacing delays
z+
z+1
Original
Frequency
8000
0.8
A = 0.6
6000
4000
2000
0
0.6
0.5
1.5
Time
Warped LPC resynth,
0.4
2.5
2.5
= -0.2
A = -0.6
0
0
0.2
0.4
0.6
0.8
^
W
Frequency
8000
0.2
6000
4000
2000
0
0.5
1.5
2
Time
http://www.ee.columbia.edu/~dpwe/resources/matlab/polewarp/
E4896 Music Signal Processing (Dan Ellis)
2013-02-25 - 14/16
Cross-Synthesis
2013-02-25 - 15/16
Summary
2013-02-25 - 16/16