![](https://csdnimg.cn/release/download_crawler_static/1950906/bg1.jpg)
Joumal
of
SoutIleast
UnjVcrsity(English
Edition)
V01.24,No.1,pp.46—49
Mar.2008
ISSN
1003—_7985
Application
of
waVelet
package
filtering
in
the
de—noising
of
fiber
optic
gyroscopes
Wang
Qi
Xu
Xiaosu
(Sch00l
of
Instnlment
Science卸d
Enginee咖g,SoucheaSt
UrIiVersity,N删ing
210096,Chjna)
Abstr习Ict:To
reduce
t11e曲ft
e盯or
existing
in山e
output
signal
of
flber
optic
gymscopes(FC)G),a
mathematical
model
of
the
F()G
outDut
signal
is
set
up;ttle
error
characteristics
of
nle
Fl×;output
signal
are
analyzed,
卸d
semi.soR
mreshold
6lte^ng
is
chosen
based
on
me
comDarison
of
h砌threshold粕d
soft
threshold
filterin2.The
senli.soft
threshold
wavelet
packa2e
filtering
method
is
applied
in
the
filtering
of
tIIe
F()G
0utput
sigllal.
ExpeTiments
of
the
stationary觚d
dynalllic
R,‘;output
signals
filtered
witll
ttle
wavelet
packa2e
allalVsis
are
car—ed
out
in
a
lab
environment,
respectively.
Experirnents
done
with
the
real—time
meaSured
FoG
si印al
show
t11at
tlle
metllod
of
serIli.son
threshold
wavelet
packa2e
fjltering
reduces
the
mean
square朗Tor
f}啪5
(。)/h
t0
1(o)/h,∞it
is
ef‰tive
in
eliminating
t11e
whitc
noises
and
t11e
fractal
noi∞s
existin£in
t11e
FoG.111e
novel
me蝴propo鼬d
here
is口roved
valid
in
reducin2Ⅱle
FoG鲥ft
e盯or'satisfying山e
tecⅢcal
deInands
of
high
precision
and
real—
time
processing.
Key
words:waVelet
pacl【age锄alysis;sjgnal
processing;fiber
optic
P吖ro;threshold
filtering
T
t
is
imponant
to
deteTllline
tlle
c枷er北imuth
angle
a11d
1
the
attitude卸gle
accurately卸d
quickly
in
a
strapdown
inertial
navi2ation
system.Gyr0
drift
js
tlle
most
imponant
factor
that affects
tlle
accuracy
of
a
s仃apdown
inertial
navi.
gation
system.How
to
eliminate
tIle
gyro
drift
eff&tively
is
an
essential
problem
tO
guarantee
the
strapdown
inertial
nav.
igation
system accuracy….The
gyro
drift
is
classified
into
svstematic嘶ft
and
random
driR.The
svstematic
driR
can
bc
eliminated
bv
a
satisfactorv
matllematical
model锄d
drift
compensatiOn
calculation.The
random
drift
cOmposed
of
white
noises柚d
fractal
noises
is
weak
nonlinear.slow
time
invariant
aIld
is
often
af琵cted
bv
some
indetenlljnate
factors
such
as
exterior
environmental
noise.Traditional
methods
t0
comDensate
for
me
random
drift
are
nOt
effective
in
the
strapdown
inertial
naVigation
system.
There
is
much
colored
noise
in
the
gyro
random
drift.
such豁f}actal
noises
characte^zed
by
non—stationary,long—
tem
dependence
and
self.simil撕ty.If
a
traditional
filter
is
used
to
process
t11e
noise,the
complexity
will
increase卸d
other
errors
will
be introduced
simultaIleousIv.
Wavelet
analysis
is
especially
suitable
for
non—statiOnaIy
fbctal
signal
R∞eived
2()07.06旬8.
Biog糟pM嚣:W衄g
Qi(198l一).male,graduate;Xu)(iaosu(co仃esponding
卸ttlor),male,doctor。pmfessor,xxs@∞u.。du.cn.
Fo帅da60n
ite咖:PI℃.Re鸵arch
PID盯am
of
General
A皿n帅ent
Depanment
during山e
11£ll
Five.Ye缸PIan
Period(No.5
l
309020503),t|le
National
De.
fen∞B勰ic
Re蛐h
Prog瑚l
ofChina(973
Prog锄)(No.973—61334),llle
NalionaJ
N咖疆l
Sci朗ce
F0unda“on
of
china(No.50575042),Specialized
Research
Fund
fbf
the
D0ctoml
Prog姗of
Higher
Education(№.
200502860261.
Cita60Ⅱ:w姐g
Qi,XuⅪ∞su.Application
of
wavelet
package
filteIing
in
山e
de。no嘲ng
of
fibcr
o曲c
gyroscopes【J】.Joumal
of
sou山east
UniVersity
(Engnsh
Edition),2008,24(1):46—49.
processing
a11d
with
relatiVely
good
fjltering
results.In
this
paper,the
wavelet
package仃aJlsfo咖is
utilized
to
perfonTl
multi-resolution卸alysis
in
gyro
signals.(bod
filtering
re—
sults
are
obtained
by
tlle
waVelet
package
filtedng
method
which
has
already
been
applied
in
pmctical
engineering
sys—
temS
1
WaVelet
Analysis
of
FOG
signal
Time.domain
presentation
Of
the
FoG
signal,
in
which
fbm
the
signals
are
usually
provided,
does
nOt
show
the
signal
propenies
except
for their
v撕ations
Over
time.On
tlle
other
hand,fl_equency
representations
of
the
si2nal
shOws
diff色rent
ftequency
components
existing
in
ttle
signal,inclu.
ding
me
vehicle
dynamics,white
and
colored
noises,and
10n2.tem
as
well
as
short.tenn
erIors.To
overcOme
the
shortcomings
of
the
FOurier仃ansfbn_11
and
the
inverse
Fou.
rier
t啪sfb咖,the
wavelet
transfbrm(WT)analyzing
the
lo.
calized副.ea
of
a
large
signal
with
a
v耐able.sized
window
is
adopted
in山e
analysis
of
the
FOG
signal.TheⅥ叮is
capa.
ble
of
providing
me
time柚d
f诧quency
infornlation
simulta.
neously,thus,giVing
a
tjme—fkquency
representatiOn
Of
the
signal.StI.etching
or
compressing
me
window
widtll
is
re.
ferred
t0
as
scaling
a
wavelet.Long
time
interval
windows
are
used
where
a
precise
low
f}equencV
component
is
needed
觚d
short
intervals
where
high
fbquency
infonnation
is
con.
sidered.Usually,t量le
low
f}equency
contents
of
tIle
signal
are
the
most
imponant
parts
of
it
that
identify
the
signal
trend
aIld
are
c印able
of
proVldlng
Very
good
印proxlmatlons
about
ttle
signal.The
approximations
correspond
to
the
high
scale
low
f诧quency
part.0n
the
other
hand,t11e
high
fre—
quency
contents
correspond
t0
the
low
scale
high
f诧quency
pan.Therefore,a
wavelet
multi.resOlution
analysis
is
per-
fonTled
based
on
me
approximation
a11d
details
provided
u.
sing
the、Ⅳ1'.The
wavelet
multi—resolution
analysis
decom.
poses
t11e
FOG
signals
intO
bottl
coarse
resolution,
which
contains
infbnIlation
about
low
f}equency
components
and
retains
the
main
features
of
tlle
original
signal,and
fine
reso—
lution
witll
infomation
about
the
high
f把quency
compo—
nents.FurthenIlOre,this
process
is
reversibIe
aS
the
signal
re—
constmction
is
defined
as
the
summation
Of
the
final印pmx.
imatjon
components
and山e
detail
componentS
of
all
levels.
1.1
WaveIet
tm眦fom
The
waVelet
transfonTl
was
OriginaIly
used
to
appmximate
a
signaJ
Of
a
function
using
a
set
Of
fUnctions,ca儿ed
a
waVe·
let
basis,which
is
proVided
by
the
mother
waVelet.WaVelets
are
a伽[Tlily
of
basis
or
baSis—like
functions
fonlled
by
dila-
tion
and仃anslation
of山e
mo山er
waVelet
function沙(工):
‰心,专(等)
㈩
万方数据