The automatic method to determine a optimal constant of wavelet threshold function of minFDR for...

download The automatic method to determine a optimal constant of wavelet threshold function of minFDR for mitigating interference application to industrial systems

of 6

Transcript of The automatic method to determine a optimal constant of wavelet threshold function of minFDR for...

  • 7/31/2019 The automatic method to determine a optimal constant of wavelet threshold function of minFDR for mitigating int

    1/6

    VCCA-2011

    VCCA-2011

    The automatic method to determine a optimal constant of wavelet threshold

    function of minFDR for mitigating interference application

    to industrial systems

    (1) (2)(1)ng [email protected]

    (2)Vin Nghin c n t, Tin hc, Tng Ha, H Ni, [email protected]

    tr da tr wat tr t tr d rta tr t trong phn t t k p t rt

    a a ra t ppp t att tr d t t t pt ta dr rat- ak tr t

    pAbstractThere are 3 steps in de-noise by wavelet

    thresholding. The second step is very

    important. It determines threshold value and

    using threshold function to mitigates noise

    coefficients. The exactitude and confidence

    in the analysis, appraise of signal is

    dependent on this step. The paper proposes a

    automatic method to determine a optimal

    constant of wavelet threshold function using

    minimize of the false discovery rate ruleminFDR for mitigating interference

    application to industrial systems.

    K hi aw atw at tr (.) x x k p

    1W a tr at t a a

    k t a

    i t a

    )(mp t t t

    N d t L1 t atL2 d NL 22 log

    h, g a at 1(.) Hm phn b tch lu ca phn b chuna t tr

    C SNR Signal Noise Rate

    DWT Decrete Wavelet Transform

    IDWT Inverse DWT

    TIDWT Translation Invariant DWT

    1. r at s c gi tr ln mangnhiu thng tin v tn hiu (thng tin hu ch) cn cch s nh mang t thng tin c ch m ch yu linquan ti nhi ta tng s dng thut loi b cc h s mang nhi a t toli tn hiu gc bng bi c wavelet.P pp ng wavelet thc hin qua a c a :

    -c mt l bi i thu at ccc h s bi i wavelet.

    - a nh gi trng v s dng c b cc h s nhiu.

    -c ba l bi at khi phctn hiu.Ca u quan trng trong qu trnh x l tnhiu. x tin cy trong phn tch, u qu kh nhiu phthuc rt nhi bc hai. u tin trong kh nhiu l chn wavelet chobi i thu c wavelet. Sau khi bi ithu at c cc h s w, thc hin giithut c cc gi tr ng w :

    )/( ww (1)

    hoc: )( ww

    , (2)

    tr l gi trng, ng v ng ca lch chun nhiu. Cc bin

    62

    mailto:[email protected]:[email protected]
  • 7/31/2019 The automatic method to determine a optimal constant of wavelet threshold function of minFDR for mitigating int

    2/6

    VCCA-2011

    VCCA-2011

    php kh nhiu khc nhau vic chn (.), v . ng s quyt nh vic p d ivi tn hi t ng nhiu quyt nhmc nhiu v gi tr nh t cc lut

    c cui cng ca kh nhiu l bi ic wavelet cho cc h s w k p thiu:

    wWx 1 (3)

    2. Ph -N [1]

    tr : )4(log2 N

    p d t (1). N t t t t p t garit trong tr nh b p t

    t t k- SURE[2] tr t ra t a k t t t a p N t p t tr t :

    ),(minarg0 ll

    wSURE (5)

    wl l c at tt SURE(,wl) a k t P pp k p a a tr a k t - Bayes [3]

    N t t p :

    s

    l

    2

    , (6)

    r 2 p a 2s

    l

    p a t t :

    )0,max( 22 xs

    (7)

    tr

    lN

    nnlx

    wN 1

    22 1 (8)

    p a a d Nl

    wavelet wn p 22

    x th

    :1 = max(wn) (9)

    tt k t t 2).- minFDR[4]

    N tr k a t a tr t t d t t tr N at n tr t t tr p:

    )]/(1[2 wpn (10)

    Sa p p tr pn:

    )(...)2()1( Nppp (11)

    Bt = :

    qN

    nnp )( (12)

    a:

    2

    )(1 1

    mp (13)

    N k a t t P k ta p d [4], [5], [6], t t p t t a t a at t tr 7 t k t a k t a, t tr = :

    )14(2

    )(

    11

    .

    .

    ).1.(

    mp

    ia

    r : tr t c tnh

    phn b nhi k t a tr i

    p a t a nh t matr a at t a a a tr p t a t tr 7 a y gi tr bng 2.Trong bi ny chng ti s trnh by c th tt t t s a ny. Gi tra c xc nhvi m t kt tt t t

    p t d t t a:

    - : at .

    - : p t t t t tn hi t M (3,4,5,6,7,8).

    - 3: t k t N- 4: d - : a k

    a t tr tr tr SN rt tp

    - 6: d t tr p p a

    - 7: tr t - 8: SNR tr t

    SN a (4). tr - : a t k

    t tr t t (3, t M tn hiu.

    - : tp t t t a at k, a a t t .

    - : a tt L*M pc a SN t.P pp a tt t t tr t

    at tr t tr ra a:

    63

  • 7/31/2019 The automatic method to determine a optimal constant of wavelet threshold function of minFDR for mitigating int

    3/6

    VCCA-2011

    VCCA-2011

    t at

    - To tn hiu v nhiu kch tc N- Chn b l k tc L1

    - Tnh cc h s lc: h, g- c dch vng L2

    i = 0

    Dch vng phi cc mu tn hiu

    - ng thch nghi

    2

    )(11..).1.(

    mp

    ia

    W c cc h s wk

    a=1:1:10

    Ch ng

    0

    -Tnh cc gi tr hp bi p a

    IDWT khi phc tn hiu

    Dch vng tri cc mu tn hiu khi phc

    i=i+1

    i = L2 ?

    1

    Ly trung bnh tn hiu

    S SN SN

    Kt thc

    1

    0

    p p SNR SN

    p t : M tn hiu

    - x

    a SNR max

    64

  • 7/31/2019 The automatic method to determine a optimal constant of wavelet threshold function of minFDR for mitigating int

    4/6

    VCCA-2011

    VCCA-2011

    3. t t a a tr k t u. Ba t t Doppler t

    tr t ata, t k ta t t tr tr pka t tha t k a t t t t t. at k ad t k p ta d t pt a pp at at kt k a t t a= t t k t SN k - u Doppler

    0 100 200 300 400 500

    -1.5

    -1

    -0.5

    0

    0.5

    1

    1.5

    Tin hieu goc

    -

    0 100 200 300 400 500

    -1.5

    -1

    -0.5

    0

    0.5

    1

    1.5

    Tin hieu co nhieu

    b-

    SNR = 9.2274 dB

    0 100 200 300 400 500

    -1.5

    -1

    -0.5

    0

    0.5

    1

    1.5

    Tin hieu khu nhieu DWT

    -

    SNR = 13.3442dB

    0 100 200 300 400 500

    -1.5

    -1

    -0.5

    0

    0.5

    1

    1.5

    Tin hieu khu nhieu nguong thich nghi minFDR a=3

    -

    SNR = 13.3701dB

    0 100 200 300 400 500

    -1.5

    -1

    -0.5

    0

    0.5

    1

    1.5

    Tin hieu khu nhieu nguong thich nghi minFDR a=4

    -

    SNR = 10.9215dB

    0 100 200 300 400 500

    -1.5

    -1

    -0.5

    0

    0.5

    1

    1.5

    Tin hieu khu nhieu nguong thich nghi minFDR a=6

    -

    SNR = 9.3289dB

    0 100 200 300 400 500

    -1.5

    -1

    -0.5

    0

    0.5

    1

    1.5

    Tin hieu khu nhieu nguong thich nghi minFDR a=2

    -

    SNR = 14.7669dBt t k tr t ppr t a= t SN a t k tt t k- i

    65

  • 7/31/2019 The automatic method to determine a optimal constant of wavelet threshold function of minFDR for mitigating int

    5/6

    VCCA-2011

    VCCA-2011

    0 50 0 1 0 00 1 50 0 20 0 0 2 5 00 3 00 0 3 50 0 40 0 0 4 5 0010 0

    20 0

    30 0

    40 0

    50 0

    60 04320 mau tin hieu dien

    H3 -

    0 100 200 300 400 500 600100

    150

    200

    250

    300

    350

    400

    450

    500

    Tin hieu Dien tron nhieu

    H3b

    SNR = 24.6001dB

    0 100 200 300 400 500 60100

    150

    200

    250

    300

    350

    400

    450Tin hieu Dien khu nhieu nguong deu

    H3

    0 100 200 300 400 500 60050

    100

    150

    200

    250

    300

    350

    400

    450Tin hieu Dien khu nhieu nguong thich nghi minFDR a=3

    H3

    0 100 200 300 400 500 600100

    150

    200

    250

    300

    350

    400

    450Tin hieu Dien khu nhieu nguong thich nghi minFDR a=4

    H3

    a=4, SNR=26.9690dB

    0 100 200 300 400 500 60050

    100

    150

    200

    250

    300

    350

    400

    450Tin hieu Dien khu nhieu nguong thich nghi minFDR a=6

    H3f =26.9690dB

    0 100 200 300 400 500 60050

    100

    150

    200

    250

    300

    350

    400

    450Tin hieu Dien khu nhieu nguong thich nghi minFDR a=2

    H3g

    =29.9780dB

    - C

    0 0.5 1 1.5 2 2.5

    x 104

    -1

    0

    1

    2

    3

    4

    5

    622087 mau tin hieu do khi CH4

    H4 -

    0 100 200 300 400 500 6000

    0.2

    0.4

    0.6

    0.8

    1

    1.2

    1.4512 mau (3000-3511) tin hieu do khi CH4

    H4b

    SNR = 24.3218dB

    0 100 200 300 400 500 6000

    0.2

    0.4

    0.6

    0.8

    1

    1.2

    1.4Tin hieu thuc khu nhieu nguong deu

    H4

    0 100 200 300 400 500 6000

    0.2

    0.4

    0.6

    0.8

    1

    1.2

    1.4Tin hieu thuc khu nhieu nguong thich nghi minFDR a=

    H4

    66

  • 7/31/2019 The automatic method to determine a optimal constant of wavelet threshold function of minFDR for mitigating int

    6/6

    VCCA-2011

    VCCA-2011

    0 100 200 300 400 500 6000

    0.2

    0.4

    0.6

    0.8

    1

    1.2

    1.4Tin hieu thuc khu nhieu nguong thich nghi minFDR a=4

    H4

    =30.4424dB

    0 100 200 300 400 500 6000

    0.2

    0.4

    0.6

    0.8

    1

    1.2

    1.4Tin hieu thuc khu nhieu nguong thich nghi minFDR a=6

    H4f

    =27.3075dB

    0 100 200 300 400 500 6000

    0.2

    0.4

    0.6

    0.8

    1

    1.2

    1.4Tin hieu thuc khu nhieu nguong thich nghi minFDR a=2

    H4g

    =34.2258dB tr p t t t t t k kt t t t d tr t k t a= t t tt tr t SN a

    p pp tr t k

    4. r t t tt t tr p d d t d k a tr a a t k t a= tr k a tr a= t a kt

    p tr t r tt t ch s a t a= tc ca nhm tcgi xut l:

    2

    )(11..).1.(2

    mp

    i

    N p pp k

    tr t kt k tnghi wavelet minFDR ( a=) t tr tr SNR a t

    t k p tr tr da t ra k t d ra

    [1] D. Donoho, I. M. Johnstone: Ideal SpatialAdaptation by Wavelet Shrinkage, Department of

    Statistics, Stanford University, USA. April 1993.

    [2] Donoho, D.L., Johnstone, I.M.: Adapting to

    unknown smoothness via wavelet shrinkage.

    Department of statistics Stanford university, USA,

    1994.[3] Sanyam Anand1, Amitabh Sharma, Akshay

    Girdhar: Undecimated Wavelet Based New

    Threshold Method for Speckle Noise Removal in

    Ultrasound Images. IACSIT Press, Singapore

    2011.

    [4] V.V.K.D.V. Prasad, P.Siddaiah and B.PrabhakaraRao: Denoising of Biological Signals Using

    Different Wavelet Based Methods and Their

    Comparison. Asian Journal of Information

    Technology 7 (4): 146-149, 2008.

    [5] David Donoho and Jiashun Jin: Asymptotic

    minimaxity of false discovery rate thresholding

    for sparse exponential data. The Annals of

    Statistics, Institute of Mathematical Statistics,

    Vol. 34, No. 6, 2980 -3018, 2006.

    [6] Martina Pavlicova, Thomas J. Santner and Noel

    Cressie: Detecting signals in FMRI data usingpowerful FDR procedures. Statistics and Its

    Interface Volume 1, 2332, 2008.[7 r N r P

    : . t tr COMNAVI 2011, 142-147, 2011

    r t tt

    tt t t tr a t

    N a a - t tr ra a t t

    N tt k t t k p

    67