Beruflich Dokumente
Kultur Dokumente
By
Harikrishna satish.T
M.E. Electrical engg,
Jadavpur university.
1
Empirical mode decomposition
2
Why Empirical Mode Decomposition?
(Type I)
Such “Fourier modes” are of particular interest in the case of
stationary signals and linear systems.
3
However, many physical situations are known to undergo non
stationary and/or nonlinear behaviors.
(Type II)
The rationale for such a modeling is to compactly encode possible
non stationarities in a time variation of the amplitudes and
frequencies of Fourier-like modes.
4
More generally,
signals may also be generated by nonlinear systems for
which oscillations are not necessarily associated with
circular functions,
thus suggesting decompositions of the following form
(Type III)
Each of the components has to have physical and mathematical
meaning. 5
Empirical Mode Decomposition (EMD) is designed primarily
for obtaining representations of Type II or Type III
in the case of signals which are oscillatory, possibly non
stationary or generated by a nonlinear system, in some
automatic, fully data-driven way.
6
Main differences between EMD and
traditional data analysis methods
7
Basic Parts of the Empirical Mode
Decomposition
8
The starting point of EMD
is to consider oscillatory signals at the level of their local oscillations
and to formalize the idea that:
9
“signal = fast oscillations superimposed to slow oscillations”
10
Decomposition of the signal:-
11
Empirical Mode Decomposition
• Decomposition:
12
How to find one Intrinsic Mode Functions of a signal?
Sifting procedure
1. identify all extrema of x(t).
2. Interpolate the local maxima to form an upper envelope u(x).
13
Huang’s “Sifting Process”.
Residue = s(t)
IMF 1; iteration 0
I1(t) = Residue 2
i=1 1
k=1
0
while Residue not equal zero or not monotone
-1
while Ii has non-negligible local mean
U(t) = spline through local maxima of Ii -2
10 20 30 40 50 60 70 80 90 100 110 120
L(t) = spline through local minima of Ii
Av(t) = 1/2 (U(t) + L(t))
Ii(t) = Ii(t) - Av(t)
i=i+1
end
IMFk(t) = Ii(t)
Residue = Residue - IMFk
k = k+1
end
14
Huang’s “Sifting Process”.
Residue = s(t)
IMF 1; iteration 0
I1(t) = Residue 2
i=1 1
k=1
0
while Residue not equal zero or not monotone
-1
while Ii has non-negligible local mean
U(t) = spline through local maxima of Ii -2
10 20 30 40 50 60 70 80 90 100 110 120
L(t) = spline through local minima of Ii
Av(t) = 1/2 (U(t) + L(t))
Ii(t) = Ii(t) - Av(t)
i=i+1
end
IMFk(t) = Ii(t)
Residue = Residue - IMFk
k = k+1
end
15
Huang’s “Sifting Process”.
Residue = s(t)
IMF 1; iteration 0
I1(t) = Residue 2
i=1 1
k=1
0
while Residue not equal zero or not monotone
-1
while Ii has non-negligible local mean
U(t) = spline through local maxima of Ii -2
10 20 30 40 50 60 70 80 90 100 110 120
L(t) = spline through local minima of Ii
Av(t) = 1/2 (U(t) + L(t))
Ii(t) = Ii(t) - Av(t)
i=i+1
end
IMFk(t) = Ii(t)
Residue = Residue - IMFk
k = k+1
end
16
Huang’s “Sifting Process”.
Residue = s(t)
IMF 1; iteration 0
I1(t) = Residue 2
i=1 1
k=1
0
while Residue not equal zero or not monotone
-1
while Ii has non-negligible local mean
U(t) = spline through local maxima of Ii -2
10 20 30 40 50 60 70 80 90 100 110 120
L(t) = spline through local minima of Ii
Av(t) = 1/2 (U(t) + L(t))
Ii(t) = Ii(t) - Av(t)
i=i+1
end
IMFk(t) = Ii(t)
Residue = Residue - IMFk
k = k+1
end
17
Huang’s “Sifting Process”.
Residue = s(t)
IMF 1; iteration 0
I1(t) = Residue 2
i=1 1
k=1
0
while Residue not equal zero or not monotone
-1
while Ii has non-negligible local mean
U(t) = spline through local maxima of Ii -2
10 20 30 40 50 60 70 80 90 100 110 120
L(t) = spline through local minima of Ii
Av(t) = 1/2 (U(t) + L(t))
Ii(t) = Ii(t) - Av(t)
i=i+1
end
IMFk(t) = Ii(t)
Residue = Residue - IMFk
k = k+1
end
18
Huang’s “Sifting Process”.
Residue = s(t)
IMF 1; iteration 0
I1(t) = Residue 2
i=1 1
k=1
0
while Residue not equal zero or not monotone
-1
while Ii has non-negligible local mean
U(t) = spline through local maxima of Ii -2
10 20 30 40 50 60 70 80 90 100 110 120
L(t) = spline through local minima of Ii
Av(t) = 1/2 (U(t) + L(t))
Ii(t) = Ii(t) - Av(t)
i=i+1
end
IMFk(t) = Ii(t)
Residue = Residue - IMFk
k = k+1
end
19
Huang’s “Sifting Process”.
Residue = s(t)
IMF 1; iteration 0
I1(t) = Residue 2
i=1 1
k=1
0
while Residue not equal zero or not monotone
-1
while Ii has non-negligible local mean
U(t) = spline through local maxima of Ii -2
10 20 30 40 50 60 70 80 90 100 110 120
L(t) = spline through local minima of Ii
residue
Av(t) = 1/2 (U(t) + L(t))
1.5
Ii(t) = Ii(t) - Av(t) (“residue”-->) 1
i=i+1 0.5
0
end -0.5
IMFk(t) = Ii(t) -1
-1.5
Residue = Residue - IMFk
10 20 30 40 50 60 70 80 90 100 110 120
k = k+1
end
20
Huang’s “Sifting Process”.
Residue = s(t)
IMF 1; iteration 1
I1(t) = Residue
1.5
i=1 1
0.5
k=1
0
while Residue not equal zero or not monotone -0.5
i=i+1 0.5
0
end -0.5
IMFk(t) = Ii(t) -1
-1.5
Residue = Residue - IMFk
10 20 30 40 50 60 70 80 90 100 110 120
k = k+1
end
21
Huang’s “Sifting Process”.
Residue = s(t)
IMF 1; iteration 1
I1(t) = Residue
1.5
i=1 1
0.5
k=1
0
while Residue not equal zero or not monotone -0.5
i=i+1 0.5
0
end -0.5
IMFk(t) = Ii(t) -1
-1.5
Residue = Residue - IMFk
10 20 30 40 50 60 70 80 90 100 110 120
k = k+1
end
22
Huang’s “Sifting Process”.
Residue = s(t)
IMF 1; iteration 1
I1(t) = Residue
1.5
i=1 1
0.5
k=1
0
while Residue not equal zero or not monotone -0.5
i=i+1 0.5
0
end -0.5
IMFk(t) = Ii(t) -1
-1.5
Residue = Residue - IMFk
10 20 30 40 50 60 70 80 90 100 110 120
k = k+1
end
23
Huang’s “Sifting Process”.
Residue = s(t)
IMF 1; iteration 1
I1(t) = Residue
1.5
i=1 1
0.5
k=1
0
while Residue not equal zero or not monotone -0.5
i=i+1 0.5
0
end -0.5
IMFk(t) = Ii(t) -1
-1.5
Residue = Residue - IMFk
10 20 30 40 50 60 70 80 90 100 110 120
k = k+1
end
24
Huang’s “Sifting Process”.
Residue = s(t)
IMF 1; iteration 1
I1(t) = Residue
1.5
i=1 1
0.5
k=1
0
while Residue not equal zero or not monotone -0.5
i=i+1 0.5
0
end -0.5
IMFk(t) = Ii(t) -1
-1.5
Residue = Residue - IMFk
10 20 30 40 50 60 70 80 90 100 110 120
k = k+1
end
25
Huang’s “Sifting Process”.
Residue = s(t)
IMF 1; iteration 1
I1(t) = Residue
1.5
i=1 1
0.5
k=1
0
while Residue not equal zero or not monotone -0.5
i=i+1 0.5
0
end -0.5
IMFk(t) = Ii(t) -1
-1.5
Residue = Residue - IMFk
10 20 30 40 50 60 70 80 90 100 110 120
k = k+1
end
26
Huang’s “Sifting Process”.
Residue = s(t)
IMF 1; iteration 1
I1(t) = Residue
1.5
i=1 1
0.5
k=1
0
while Residue not equal zero or not monotone -0.5
0.5
i=i+1
0
end
-0.5
IMFk(t) = Ii(t) -1
27
Huang’s “Sifting Process”.
Residue = s(t)
IMF 1; iteration 2
I1(t) = Residue 1.5
1
i=1
0.5
k=1
0
while Residue not equal zero or not monotone
-0.5
while Ii has non-negligible local mean -1
0.5
i=i+1
0
end
-0.5
IMFk(t) = Ii(t) -1
28
Huang’s “Sifting Process”.
Residue = s(t)
IMF 1; iteration 2
I1(t) = Residue 1.5
1
i=1
0.5
k=1
0
while Residue not equal zero or not monotone
-0.5
while Ii has non-negligible local mean -1
0.5
i=i+1
0
end
-0.5
IMFk(t) = Ii(t) -1
29
Huang’s “Sifting Process”.
Residue = s(t)
IMF 1; iteration 2
I1(t) = Residue 1.5
1
i=1
0.5
k=1
0
while Residue not equal zero or not monotone
-0.5
while Ii has non-negligible local mean -1
0.5
i=i+1
0
end
-0.5
IMFk(t) = Ii(t) -1
30
Huang’s “Sifting Process”.
Residue = s(t)
IMF 1; iteration 2
I1(t) = Residue 1.5
1
i=1
0.5
k=1
0
while Residue not equal zero or not monotone
-0.5
while Ii has non-negligible local mean -1
0.5
i=i+1
0
end
-0.5
IMFk(t) = Ii(t) -1
31
Huang’s “Sifting Process”.
Residue = s(t)
IMF 1; iteration 2
I1(t) = Residue 1.5
1
i=1
0.5
k=1
0
while Residue not equal zero or not monotone
-0.5
while Ii has non-negligible local mean -1
0.5
i=i+1
0
end
-0.5
IMFk(t) = Ii(t) -1
32
Huang’s “Sifting Process”.
Residue = s(t)
IMF 1; iteration 2
I1(t) = Residue 1.5
1
i=1
0.5
k=1
0
while Residue not equal zero or not monotone
-0.5
while Ii has non-negligible local mean -1
0.5
i=i+1
0
end
-0.5
IMFk(t) = Ii(t) -1
33
Huang’s “Sifting Process”.
Residue = s(t)
IMF 1; iteration 2
I1(t) = Residue 1.5
1
i=1
0.5
k=1
0
while Residue not equal zero or not monotone
-0.5
while Ii has non-negligible local mean -1
34
Huang’s “Sifting Process”.
Residue = s(t)
IMF 1; iteration 3
I1(t) = Residue
1
i=1
0.5
k=1
0
while Residue not equal zero or not monotone
-0.5
while Ii has non-negligible local mean
-1
U(t) = spline through local maxima of Ii
10 20 30 40 50 60 70 80 90 100 110 120
L(t) = spline through local minima of Ii
residue
Av(t) = 1/2 (U(t) + L(t))
1
Ii(t) = Ii(t) - Av(t)
0.5
i=i+1
0
end
-0.5
IMFk(t) = Ii(t)
-1
Residue = Residue - IMFk
10 20 30 40 50 60 70 80 90 100 110 120
k = k+1
end
35
Huang’s “Sifting Process”.
Residue = s(t)
IMF 1; iteration 4
I1(t) = Residue
1
i=1
0.5
k=1
0
while Residue not equal zero or not monotone
-0.5
while Ii has non-negligible local mean
-1
U(t) = spline through local maxima of Ii
10 20 30 40 50 60 70 80 90 100 110 120
L(t) = spline through local minima of Ii
residue
Av(t) = 1/2 (U(t) + L(t))
1
Ii(t) = Ii(t) - Av(t)
0.5
i=i+1
0
end
-0.5
IMFk(t) = Ii(t)
-1
Residue = Residue - IMFk
10 20 30 40 50 60 70 80 90 100 110 120
k = k+1
end
36
Huang’s “Sifting Process”.
Residue = s(t)
IMF 1; iteration 5
I1(t) = Residue
1
i=1
0.5
k=1
0
while Residue not equal zero or not monotone
-0.5
while Ii has non-negligible local mean
-1
U(t) = spline through local maxima of Ii
10 20 30 40 50 60 70 80 90 100 110 120
L(t) = spline through local minima of Ii
residue
Av(t) = 1/2 (U(t) + L(t))
1
Ii(t) = Ii(t) - Av(t)
0.5
i=i+1
0
end
-0.5
IMFk(t) = Ii(t)
-1
Residue = Residue - IMFk
10 20 30 40 50 60 70 80 90 100 110 120
k = k+1
end
37
Huang’s “Sifting Process”.
Residue = s(t)
IMF 1; iteration 6
I1(t) = Residue
1
i=1
0.5
k=1
0
while Residue not equal zero or not monotone
-0.5
while Ii has non-negligible local mean
-1
U(t) = spline through local maxima of Ii
10 20 30 40 50 60 70 80 90 100 110 120
L(t) = spline through local minima of Ii
residue
Av(t) = 1/2 (U(t) + L(t))
1
Ii(t) = Ii(t) - Av(t)
0.5
i=i+1
0
end
-0.5
IMFk(t) = Ii(t)
-1
Residue = Residue - IMFk
10 20 30 40 50 60 70 80 90 100 110 120
k = k+1
end
38
Huang’s “Sifting Process”.
Residue = s(t)
IMF 1; iteration 7
I1(t) = Residue
1
i=1
0.5
k=1
0
while Residue not equal zero or not monotone
-0.5
while Ii has non-negligible local mean
-1
U(t) = spline through local maxima of Ii
10 20 30 40 50 60 70 80 90 100 110 120
L(t) = spline through local minima of Ii
residue
Av(t) = 1/2 (U(t) + L(t))
1
Ii(t) = Ii(t) - Av(t)
0.5
i=i+1
0
end
-0.5
IMFk(t) = Ii(t)
-1
Residue = Residue - IMFk
10 20 30 40 50 60 70 80 90 100 110 120
k = k+1
end
39
Huang’s “Sifting Process”.
Residue = s(t)
IMF 1; iteration 8
I1(t) = Residue
1
i=1
0.5
k=1
0
while Residue not equal zero or not monotone
-0.5
while Ii has non-negligible local mean
-1
U(t) = spline through local maxima of Ii
10 20 30 40 50 60 70 80 90 100 110 120
L(t) = spline through local minima of Ii
residue
Av(t) = 1/2 (U(t) + L(t))
1
Ii(t) = Ii(t) - Av(t)
0.5
i=i+1
0
end
-0.5
IMFk(t) = Ii(t)
-1
Residue = Residue - IMFk
10 20 30 40 50 60 70 80 90 100 110 120
k = k+1
end
40
Huang’s “Sifting Process”.
Residue = s(t)
IMF 2; iteration 0
I1(t) = Residue 1
i=1 0.5
k=1
0
while Residue not equal zero or not monotone
-0.5
while Ii has non-negligible local mean
-1
U(t) = spline through local maxima of Ii
10 20 30 40 50 60 70 80 90 100 110 120
L(t) = spline through local minima of Ii
residue
Av(t) = 1/2 (U(t) + L(t))
1
Ii(t) = Ii(t) - Av(t)
0.5
i=i+1
0
end
-0.5
IMFk(t) = Ii(t)
Residue = Residue - IMFk -1
10 20 30 40 50 60 70 80 90 100 110 120
k = k+1
end
41
An example:-
(puts emphasis on the potentially
“nonharmonic” nature of EMD)
42
Some remarks on Huang’s EMD
No analytic definition —
The decomposition is only defined as the output of an
algorithm
43
Applications of EMD
Environmental:
Oceanography, Earthquake engineering , Water and wind
dynamics.
Industrial:
Machine monitoring and failure prediction .
44
References:-
P. FLANDRIN, P. GONCALVES, 2004 :
"Empirical Mode Decompositions as Data-Driven Wavelet-Like
Expansions,"
Int. J. of Wavelets, Multires. and Info. Proc., Vol. 2, No. 4, pp. 477-496.
ON EMPIRICAL MODE DECOMPOSITION AND ITS ALGORITHMS
Gabriel Rilling∗, Patrick Flandrin∗ and Paulo Gon¸calv`es∗∗
∗Laboratoire de Physique (UMR CNRS 5672), ´Ecole Normale Sup´erieure de Lyon
46, all´ee d’Italie 69364 Lyon Cedex 07, France.
http://perso.ens lyon.fr/patrick.flandrin/emd.html
45