Signal Information Processing

yakzephyrAI and Robotics

Nov 24, 2013 (3 years and 10 months ago)

93 views

EEE Expo 2012

School of Electrical and Electronic Engineering
-

The University of Adelaide

Slide
0



Signal Information Processing

Brian Ng

EEE Expo 2012

School of Electrical and Electronic Engineering
-

The University of Adelaide

Slide
1

Multi
-
rate
Signal Processing and Wavelets


Fundamental research on multi
-
rate
filter banks


Multi
-
resolution analysis


Rational cases


Applications in time
-
frequency analysis of signals in
numerous fields


Radar signal processing


Time
-
domain
THz
spectroscopy

EEE Expo 2012

School of Electrical and Electronic Engineering
-

The University of Adelaide

Slide
2

Information Extraction

Sparse representations


M
any real life signals afford
sparse
representations


Exploited for
reconstructions from
incomplete measurements


Application in many fields, e.g. detection, rapid
measurements

Real world signals contain much useful information


Often obscured by noise, interference …
etc


Processing techniques become critical


Current projects:


Breast cancer detection


railway bearing condition monitoring


Environmental discovery from vision



2
3



Fi
g
u
r
e

15
:
W
e
lch

p
o
w
e
r

sp
e
c
tr
a
l
d
e
n
sity

o
f
th
e

r
a
w

d
a
ta

fo
r

th
e

w
in
d
o
w

len
g
th

o
f
3
2
7
6
8

F
ig
u
re

15

shows som
e

e
x
c
it
a
ti
on a
round
16kHz
.
F
uture
investi
ga
ti
ons

the PS
D is z
oomed
a
round
16kHz

a
nd show
n in
F
ig
u
re

16
.
S
ide ba
nd
s
on e
it
he
r side
of the

e
x
c
it
e
d na
tura
l
fr
e
que
n
c
y
a
r
e

se
p
a
ra
ted

b
y
18 Hz
, whic
h is close
to t
he
B
P
F
O (
17.8
Hz
).


Fi
g
u
r
e

16
:
Z
o
o
m
e
d

PS
D a
r
o
u
n
d

1
6
k
H
z


The

c
lassic
a
l
e
nv
e
lop
a
n
a
l
y
sis

invol
ve
s
b
a
nd
-
p
a
ssi
ng

the

sig
n
a
l
for

de
mo
dulation
a
round

th
e

e
x
c
it
e
d
na
tura
l
fr
e
que
nc
y

b
y

the
de
fe
c
t
-
re
late
d
im
pulses.
How
e
ve
r,
the
main
diff
icult
y

is

de
ter
mi
ning
the
p
rope
r

b
a
nd
-
pa
ss
filter
in
g
.
S
ince

c
hoosi
ng
the
fr
e
qu
e
nc
y

b
a
nd

b
y

looki
ng
a
t
the
P
S
D
c
ould
be

mi
slea
ding

(
Antoni
2006;
Antoni
&

R
a
nda
ll

2006)
,

Antoni
a
nd
R
a
nda
ll

(2006
)

su
gg
e
sted
usin
g

Kur
togra
m
in
or
d
e
r
to
s
e
lec
t
the
most

pr
op
e
r
fr
e
que
nc
y

b
a
nd
fo
r

e
nve
lope
a
na
l
y
sis
.
Kur
to
g
r
a
m
is

the
r
e
pr
e
s
e
ntation
of

S
K
with
diff
e
re
nt
w
indow
leng
ths

for

diff
e
re
nt
fr
e
qu
e
nc
ies.

F
i
g
ur
e

17

shows
th
e

Kur
t
og
r
a
m,
whic
h

d
e
picts
th
e

c
a
l
c
ulate
d
S
K
fo
r
diff
e
re
nt windo
w le
n
g
t
hs (le
ve
ls) a
ga
inst
the r
e
lat
e
d fr
e
qu
e
nc
y
b
a
nd.

0
100
200
300
400
500
600
−1
−0.8
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
0.8
50
100
150
200
250
50
100
150
200
250
50
100
150
200
250
50
100
150
200
250