Professional Documents
Culture Documents
Fast MEMD PDF
Fast MEMD PDF
net/publication/328434279
CITATIONS READS
5 417
7 authors, including:
Hongye Su
Zhejiang University
549 PUBLICATIONS 9,336 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Lang Xun on 30 October 2018.
Abstract—The multivariate empirical mode decomposition publications on EMD has been increasing steadily over the
(MEMD) has been pioneered recently for adaptively process- past decade [5].
ing of multichannel data. Despite its high efficiency on time- In recent years, advances in signal acquisition tools have
frequency analysis of nonlinear and nonstationary signals, high
computational load and over-decomposition have restricted wider highlighted the requirement for synchronous processing of
applications of MEMD. To address these challenges, a fast multichannel data [6]. As a result, the standard EMD has
MEMD (FMEMD) algorithm is proposed and featured by the been extended to various versions including those suitable for
following contributions: (i) A novel concept, pseudo direction- the bivariate [7], trivariate [8] and multivariate signals [9].
independent Multivariate Intrinsic Mode Function (IMIMF) Bivariate EMD (BEMD) [7] estimates the bivariate local mean
which allows the interchange of sifting and projection operations,
is defined for the purpose of developing FMEMD, (ii) FMEMD by mapping the input to a number of real-valued projections
is computationally efficient. Compared with MEMD, the number and then averaging the respective local mean obtained from
of time-consuming sifting operations reduces from K · p to K for these projected signals. Following the same idea, ur Rehman
each iteration, where K and p denote the number of projection et al. [8] extended the EMD algorithm to Trivariate EMD
directions and signal dimension, respectively, (iii) FMEMD is (TEMD) by taking the three-dimensional signal as a pure
consistent with EMD in terms of the dyadic filter bank property
and (iv) FMEMD is more effective in working at low sampling quaternion. Similarly, the multivariate EMD (MEMD) [9]
rate. Validity of the raised approach is demonstrated on a wide further generalizes this concept for multivariate signal and
variety of real world applications. efficiently solve the uncertainty of IMFs arrangement [10].
Index Terms—Multivariate empirical mode decomposition, The MEMD performs well in many applications from
multivariate intrinsic mode function, fast MEMD, computational various fields such as process control [11], mechanical health
load, filter bank. diagnosis [10], biomedicine [12], data denoising [13], etc.
However, the major drawback of MEMD is that it requires a
long computation time especially for high-dimensional signals.
I. I NTRODUCTION
More precisely, a total number of K · p sifting operations are
from K · p to K. (iii) The filter bank property of FMEMD is B. Multivariate Empirical Mode Decomposition
analyzed with the study on white Gaussian noise (WGN). It
The standard EMD is only applicable for univariate signal.
reveals that FMEMD is consistent with EMD in the sense that
For multivariate signals, however, the local maxima and min-
it also acts as a dyadic filter bank. (iv) The proposed FMEMD
ima cannot be defined directly, and the notion of oscillatory
avoids the over-decomposition problem and is more effective
mode defining a multivariate IMF is not straightforward either
in working at low sampling rate, since the IMFs for different
[7]. To tackle this problem, multiple p-dimensional envelopes
projections are obtained independently.
are generated by taking signal projections along different
The rest of this paper is organized as follows. Traditional directions in p-dimensional space, and subsequently interpo-
EMD and MEMD are first reviewed in Section II. Section lating their extrema [9]. These envelopes are then averaged to
III details the proposed FMEMD approach. In Section IV, obtain the local multivariate mean.
the filter bank properties of FMEMD are analyzed by de- Let x (t) = [x1 (t) , x2 (t) , . . . , xp (t)] be a p-variate
composing multichannel white Gaussian noise. A comparative K
signal, and vθk = v1k , v2k , . . . , vpk k=1 denotes the kth
study between FMEMD and MEMD is presented in Section V.
projection vector alongio the direction given by angle
Section VI studies the FMEMD performance with three real n h K
world applications. It is followed by conclusions in Section θk = θk1 , θk2 , . . . , θkp−1 on an (p − 1) sphere1 , the
k=1
VII. MEMD algorithm is summarized in Algo. 2.
i=1
6: Extract the detail signal s (t) using s (t) = x1 (t) − m (t).
where the residual signal r (t) represents the trend within the If the detail signal satisfies the stopping criterion of MIMF,
signal. The procedures used for extraction of IMFs from a di (t) = s (t), go to step 7. Otherwise x1 (t) = s (t), go
real-valued signal x (t) are summarized in Algo. 1. to step 2;
7: Update x2 (t) = x2 (t) − di (t). If x2 (t) becomes mono-
Algorithm 1 Empirical mode decomposition. tonic, or does not contain enough extrema to form a
Input: x1 (t) = x2 (t) = x (t), i = 1 meaningful multivariate envelope, stop the above sifting
1: Find the locations of all the extrema of x1 (t); process, and obtain the trend r (t) = x2 (t). Otherwise,
2: Interpolate (typically using cubic spline interpolation) all let i = i + 1 and x1 (t) = x2 (t), go to step 2.
M
the maxima (minima) to obtain the upper (lower) enve- 8: return {di (t)}i=1 and r (t).
lope, emax (t) (emin (t));
3: Find the local mean, m (t) = [emin (t) + emax (t)]/2; As a result, MEMD uses a vector-valued form of standard
4: Subtract the mean from the signal to obtain the oscillatory EMD to decompose a p-variate signal x (t) as
mode, s (t) = x1 (t) − m (t);
M
5: If s (t) obeys the stoppage criterion, di (t) = s (t) be- X
comes an IMF, go to step 6. Otherwise set x1 (t) = s (t) x (t) = di (t) + r (t) , (3)
i=1
and repeat the procedures from step 1;
M
6: Subtract the derived IMF from x2 (t), so that x2 (t) = where the p-variate MIMFs, {di (t)}i=1 , contain scale-aligned
x2 (t) − di (t). If x2 (t) becomes a monotonic function, or intrinsic joint rotational modes [2].
does not contain enough extrema to form a meaningful
envelope, stop the sifting process with r (t) = x2 (t). 1 A (p − 1) sphere (hypersphere) is an extension of the ordinary sphere
Otherwise, x1 (t) = x2 (t), i = i + 1 and go to step 1; to an arbitrary dimension and is represented mathematically in R =
p
M P
(xi − Ci )2 , where R denotes the radius and Ci is the ith coordinate
7: return {di (t)}i=1 and r (t);
i=1
value of center point C. This work adopts the terminology that a (p − 1)
sphere resides in a p-dimensional Euclidean coordinate system.
IEEE ACCESS 3
Y
supplementary statements.
projection
direction
A. Definitions around MIMF
signal
The standard EMD is implemented based on the intuitive direction
(2) the number of zero-crossings and extrema must either fulfilled after l times sifting, a well-developed DMIMF can be
equal or differ at most by one. computed as
dθk (t) = D| ·{z
· · D} x (t) . (4)
Remarks:
l
1) The defined IMIMF is an intersection of all sets of
DMIMFs that satisfying conditions from definition 2. It It can be proved that if the standard EMD algorithm is
requires that the projection of IMIMF on each of the meaningful on the projected signal, there exists a positive
directions is a well-defined IMF. As a result, the physical integer l so that Eq. (4) meets the requirements for defining a
meaning of the univariate IMF has been extended to DMIMF.
multivariate case by IMIMF in a strict sense. The IMIMF
Lemma 1. The projection of the cubic spline interpolation
is a symmetrical multivariate-oscillation with local zero
curve of a multi-dimensional point set is equivalent to the
mean, for which the instantaneous frequency can be
interpolation that is calculated from the projected point set.
defined everywhere. Similarly, although [7], [9] have not
provided a clear definition of the rotation (MIMF), it can A mathematical proof of the proposed lemma is presented
be inferred that the projection of rotation for any given in Appendix. A.
direction is also a zero-mean oscillation. Therefore, the
defined IMIMF and the rotation defined by MEMD are Theorem 1. Given a specified direction vector, there exists
essentially the same notion. Both of them are naturally a positive integer l that makes D
| ·{z
· · D} x (t) satisfying the
extended from the definition of univariate IMF (oscilla- l
tion). On the other hand, they realized the computation conditions of DMIMF if and only if EMD is convergent on
of MIMF/IMIMF in different ways, and the next few the projected signal.
sections will show that the proposed IMIMF is a more Proof. Let x (t) denotes the investigated multivariate signal,
effective extension. xθk (t) represents the projection of x (t) along vθk , the rela-
2) Extending the sifting process for computing a standard tionship between x (t) and xθk (t) is given as
IMF to IMIMF calculation is unrealizable in practice
due to following difficulties: (i) Finding all multivariate x (t) · vθk = xθk (t) . (5)
extrema requires projecting the original signal onto infi-
nite directions; (ii) Determining whether a given signal is According to definition 1, the extracted multivariate extrema
IMIMF also needs to check its projections onto infinite of x (t) and the univariate extrema of xθk (t) fulfill the
directions. following equation set
3) A pseudo-IMIMF that defined only on K directions will
max θ
xm · v k = xmax m , m = 1, 2, . . . , M
be given in Section III-C as an alternative, where K > p . (6)
xmin
n · v θk
= x min
n , n = 1, 2, . . . , N
is a finite positive integer. min N
M
where {xmaxm }m=1 and xn n=1 denote the multivariate
max M
N
B. Interchangeability for Computing DMIMF extrema, {xm }m=1 and xmin n n=1
are univariate extrema
of xθk (t) respectively. Lemma 1 shows that projection of
Before addressing the problem of calculating (pseudo) IM-
the cubic spline interpolation curve of a multi-dimensional
IMF for the purpose of FMEMD development, the direction-
point set is equivalent to the interpolation calculated from the
dependent MIMF (DMIMF) is first computed in combination
projected point set. Therefore it can be concluded that
with a theoretical demonstration of the interchangeability
S (xmax ) · vθk = S xmax θk max
between (multivariate) sifting process and direction-base pro- m m · v =S (xm )
min θk min θk min , (7)
jection. Given a specified direction vector vθk , the sifting S xn · v = S xn · v =S xn
process of standard EMD (step 1 to step 4 in Algo. 1) can
where S(·) denotes the spline interpolation operator, which
be extended multivariable for decomposing DMIMF from a
yields
p-variate signal x (t), as given by:
eu (t) · vθk = eθuk (t)
1) Find the locations of all maxima {xmax
M , (8)
min N m }m=1 and minima el (t) · vθk = eθl k (t)
xn n=1 on direction vθk .
2) Interpolate (using cubic spline interpolation channel-wise in which eu (t) (eθuk (t)) and el (t) (eθl k (t)) represent the upper
[9]) all the maxima (minima) to obtain the upper (lower) and lower envelopes of x (t) (xθk (t)), respectively. On the
envelope, eu (t) (el (t)). basis of Eq. (8), the relationship between the local mean
3) Compute the local mean, m (t) = [eu (t) + el (t)]/2. (m (t)) of x (t) and the local mean (mθk (t)) of xθk (t) is
4) Subtract the mean from the signal to obtain the oscillatory given by
modes, o (t) = x (t) − m (t). [eu (t) + el (t)] θk
5) If o (t) obeys the two conditions in definition 2, a m (t) · vθk = ·v
2
DMIMF is obtained. Otherwise, repeat step 1 to 4 on
h i . (9)
eθuk (t) + eθl k (t)
o (t). = θk
= m (t)
Let D denotes the operation from step 1 to 4, then Dx (t) = 2
x (t) −m (t). By supposing that conditions for definition 2 are As a result, the so obtained oscillatory modes Dx (t) and
IEEE ACCESS 5
Dxθk (t) show the following relationship extraction’ satisfy interchangeability. Therefore, the ideal
θk θk direction-independent MIMF (IMIMF) as an intersection of
Dx (t) · v = [x (t) −m (t)] · v
. (10) all meaningful DMIMFs should also meet the same property.
= xθk (t) −mθk (t) = Dxθk (t) Mathematically, it requires the theoretical IMIMF id (t) of the
Eq. (10) proves that operations ’multivariate sifting + analyzed signal x (t) satisfies the following equation
projection’ and ’projection + univariate sifting’ satisfy in- K
id (t) · vθk =dθk (t) · vθk = dθk (t) k=1 , K → ∞, (14)
terchangeability. This useful property can be exploited for
K
FMEMD development, which will be further discussed in
where dθk (t) k=1 denote all sets of DMIMFs calculated
Section III-C. K K
upon direction vectors vθk k=1 , and dθk (t) k=1 are cor-
Since the operator D does not change the projection rela-
responding IMFs extracted from the projected signals. Since
tionship, a new equation is obtained further
K is an infinite positive integer, the above system of overde-
termined linear equations is unsolvable in engineering. To
D · · · D x (t) · vθk = D · · · D xθk (t) . (11) address the problem, a pseudo definition of IMIMF with
| {z } | {z }
l l relaxed constraint is presented by limiting K into a finite
positive integer, which is,
Part I: If the EMD is not convergent on the projected
K
signal, none of the IMFs can be extracted from xθk (t). Definition 4. For any given set of direction vectors vθk k=1 ,
Therefore, it is impossible to find such a positive integer l where K is a limited and positive integer. The pseudo-IMIMF
that causes D · · D} xθk (t) to be a proper IMF. In accordance
| ·{z id (t) of a multivariate signal satisfies the following equation:
l K
id (t) · vθk =dθk (t) k=1 ,
with definition 2 and Eq. (4), dθk (t) does not exist due to the (15)
infeasibility of l. θ K
Part II: If the EMD algorithm is meaningful on the projected where d k (t) k=1 are corresponding IMFs extracted from
signal, the projected signal must yield an univariate IMF after the projected signals..
integer times sifting. By supposing it costs l0 times sifting for If a p-variate signal x (t) is projected into K directions
decomposing the current IMF (dθk (t)) from xθk (t), we have to yield K projections (K ≥ p, and it is generally set as
dθk (t) = D · · D} xθk (t) . K > p for reducing noise), a set of overdetermined linear
| ·{z (12)
equations can be obtained after the corresponding univariate
l0
IMFs being extracted. Accordingly, the first pseudo-IMIMF
Combine Eq. (11), Eq. (12) with definition 2, the DMIMF can be computed by solving Eq. (15), and the first slower
dθk (t) that defined on vector vθk can be properly addressed mean, s (t), is therefore subtracted into
after l0 times multivariate sifting, which satisfies the following
relationship s (t) =x (t) − id (t) . (16)
The procedures can be re-applied by setting the mean signal
dθk (t) · vθk = D · · D} x (t) · vθk
| ·{z
s (t) as a new input, then the iteration continues until a
l0 . (13) monotonic trend r (t) is obtained. In consequence, an indi-
=D · · D} xθk (t) =dθk (t) rect multivariate extension of the EMD algorithm (namely,
| ·{z FMEMD) is realized. For convenience, the rest of this paper
l0
uses IMIMF instead of the defined pseudo-IMIMF.
This equality proposes a complete scheme for computing
the DMIMF on such specified direction. The projective rela-
tionship between the calculated DMIMF and the univariate D. Algorithm of FMEMD
IMF extracted from the projected signal has been proved Definition 4 enables FMEMD to be generalized indirectly
theoretically. by making full use of the decomposed results from the
univariate EMD. Based on solving an overdetermined system
The presented theorem 1 also shows the irrationality of the of linear equations (Eq. (15)), the p-variate IMIMF of the input
original MEMD method [9]. For each specified direction vec- data can be simply computed, which yields the slower mean
tor, the T-F feature of the current projection is different from as: mean = signal - IMIMF (Eq. (16)).
other projected signals, therefore, distinct numbers of itera- Consider a sequence of p-dimensional signal x (t) =
tions are required for extracting different direction-dependent [x1 (t) , x2 (t) , . . . , xp (t)] which represents a multivariate sig-
MIMFs. As a result, running of the original MEMD, which K
nal with p components, and vθk = v1k , v2k , . . . , vpk k=1
enforces the same number of sifting for all data channels, may
denote ha set of vectors alongio the directions given by angles
induce the effect of over-decomposition. n K
θk = θk1 , θk2 , . . . , θkp−1 on a (p − 1) sphere, the pro-
k=1
posed FMEMD decomposes signal x (t) into
C. Raise of Pseudo-IMIMF
M
Inspired from theorem 1, it is observed that processes X
x (t) = idi (t) + r (t) , (17)
’DMIMF extraction + projection’ and ’projection + IMF
i=1
IEEE ACCESS 6
M
where the p-variate IMIMFs, {idi (t)}i=1 , contain scale- The Halton and Hammersley sequence, however, does not
aligned intrinsic joint components. The algorithm of FMEMD usually provide suitable samplings. Consider that a direction
T
is summarized in Algo. 3. vector [1, 0] is selected to process a bivariate signal, x (t) =
[sin (3πt) , cos (6πt)], then the univariate IMF extracted from
Algorithm 3 Algorithm of FMEMD. the projected function would be sin (3πt). More specifically,
Input: x1 (t) = x (t), i = 1 the direction vector with a zero ingredient obscures one
1: Generate a suitable point set for uniform projection; component of the bivariate input, which causes the higher
2: Calculate the kth projection q θk (t) of the input signal frequency component x2 (t) = cos (6πt) to be projected into
x1 (t) along direction vector vθk , for all k (i.e. k = zero. Therefore, a vector whose direction is very close to an
1, 2, . . . , K, where K gives the whole set of vectors); unit coordinate vector will be harmful to the estimation of
3: Extract the first univariate IMF dθk (t) of the projected IMIMF. To resolve this issue, cosine similarity is introduced
function q θk (t) for all k using the standard EMD algo- for further filtering of the invalid vectors. Given two vectors
rithm; of attributes, A and B, the cosine similarity, cos(θ), is
K
4: Combine all IMFs dθk (t) k=1 with their corresponding represented using a dot product and magnitude as
θ K
direction vectors v k k=1 , the p-variate IMIMF idi (t) n
P
ai bi
can be obtained by solving the following overdetermined A·B
equations cos (θ) = = s i=1 s , (19)
kAk2 kBk2 n n
a2i b2i
P P
θ K K
d k (t) k=1 = idi (t) · vθk k=1 . (18) i=1 i=1
to the normal equations as given by Averaged spectra of white noise realizations from FMEMD
−1 IMIMF 8 7 6 5 4 3 2 1
xo = ST S ST y,
(22)
Spectrum
where xo is the unique optimal solution to x. To guarantee the
existence of ST S, a mechanism to determine whether ST S is
full rank is preferentially required. If such condition cannot 0 1 2
10 10 10
be satisfied, the starting point of Halton and Hammersley Frequency(Log)
Zero crossings(log 2)
sequence should be changed by a new random integer to obtain 10
FMEMD approximate slope = −1.0104
a new set of direction vectors. Actually, the constructed matrix 8
2
1 2 3 4 5 6 7 8
IV. F ILTER BANK P ROPERTY A NALYSIS IMIMF index
−30
In addition, the noise-based filter bank property analysis of
−40
MEMD is conducted, which is dedicated to confirm that
−50 FMEMD acts more similar to the standard EMD than that
−60 0
of MEMD. Finally, a low-sampling rate signal is synthesized
1 2 3
10 10 10 10
Frequency to compare them further. The Matlab codes of MEMD and
EMD for developing the proposed FMEMD are available
online4 . Throughout this work, the Halton and Hammersley
Fig. 3. Fourier spectra of IMIMFs from the proposed FMEMD with modes sequence is used for generating a set of K = 64 direction
3 − 8 shifted so as to overlap with the spectrum associated with mode 2. vectors. The stoppage criterion is S = 5, and end effects
of all investigated signals are restrained in advance before
FMEMD/MEMD being applied.
where m denotes the slope, c represents the y-intercept of the
straight line and is dependent on the length of the data been
investigated. Accordingly, the base-2 exponent of Eq. (24) is A. Computational Cost Analysis
given below
In this section, a simple comparison between the compu-
z [k] = 2mk · 2c = 2−m·(−k) · 2c . (25) tational loads of MEMD and FMEMD is presented. It is
noteworthy that some other fixed-point arithmetic operations
In this simulation, the approximated slope is m = −1.0104.
are negligible compared with the sifting procedure5 . Similarly,
Combining the above equation with Eq. (23), parameter ρ can
the infrequent matrix operation, Eq. (22), which is utilized for
be therefore computed as
the solution of IMIMF, also has very small effect on the total
ρ = 2−m = 21.0104 = 2.0145. (26) computational time due to two reasons:
−1
i) The time-consuming operation ST S in Eq. (22)
This regression reveals that the proposed FMEMD is em-
requires computing the inverse of a p × p matrix, ST S. In
pirically a dyadic filter capable of separating the WGN into
contrast, the computation of cubic spline interpolation needs
IMIMF components having mean frequency exactly half value
to solve the inverse of a (n − 1)×(n − 1) matrix, Q (Eq. (45)),
of the previous components.
where n denotes the number of maxima/minima extracted from
Another important finding gained from the filter bank analy-
the input signal. It is clear that the value of n of a real-world
sis of the standard EMD is the self-similarity of its constituent
signal will far exceed the dimension of the signal p. Therefore,
bandpass filters [22], [24]. To illustrate that the power spectra
the computational time of solution xo is much smaller than
of IMIMFs obtained from the proposed FMEMD also exhibit
the time cost of a single cubic spline interpolation.
this self similar property, the averaged spectra corresponding
ii) Unlike MEMD or FMEMD that−1
repeats the cubic spline
to IMIMFs 2 − 8 are properly renormalized into a curve-
interpolation for many times, ST S ST is computed only
cluster with the same center frequency based on coefficient ρ.
once during the whole process of the FMEMD decomposition,
Fig. 3 gives the redrew curves of these spectra. It is observed
which greatly saves the computational time.
that the power spectra of all IMIMFs can be collapsed onto
a single curve, when ρ based renormalization is conducted. Based on the above statements, the sifting time based
This diagram is in agreement with the EMD-base findings comparison is deducted between MEMD and FMEMD, to
reported in [22], [23], [24]. To this end, let Pk (f ) denotes facilitate the complexity analysis.
the frequency response of the kth IMIMF, self-similarity of the Let us suppose a single operation of the univariate sifting
bandpass filters for both EMD and FMEMD can be described process for standard EMD requires T0 time. According to [9],
by the multivariate envelope associated with a specified projection
Pn (f ) = Pk ρn−k f ,
(27) is computed by applying the univariate cubic spline interpo-
lation channel-wise. Notice that the cubic spline interpolation
where n ≥ k ≥ 2. The numerical analysis in this section in a single sifting process occupies most of the computation
supports the claim the proposed FMEMD is a consistent for both EMD and MEMD. Consequently, focus on each of
generalization of the standard EMD. the direction vectors, the sifting time of MEMD of a p-variate
input is about p times of the univariate sifting time, T0 . If
V. C OMPARATIVE S TUDY BETWEEN MEMD AND FMEMD
4 The code of MEMD is obtained from: http://www.commsp.ee.ic.ac.uk/
The proposed FMEMD is a well developed extension of the
∼mandic/research/emd.htm, and the code of EMD can be found in: http://
univariate EMD for time-frequency analysis of multivariate perso.ens-lyon.fr/patrick.flandrin/emd.html.
data. Unlike traditional MEMD, the raised method is able to 5 The sifting process includes all of the large computational operations,
not only remove the dependence of the computational com- such as extrema identification and cubic spline procedures, where the cubic
spline interpolation occupies most of the computation. In contrast, the rest
plexity on dimension of the input signal, but also intuitively of MEMD/FMEMD only involves some simple operations like addition,
present a more desired decomposition result. To verify the subtraction, multiplication and division.
IEEE ACCESS 9
totally K direction vectors are considered, the time cost of the stopping criteria. However, not every channel of the real
MEMD for extracting a single MIMF is given by world signal needs to have the same number of iterations. As
a result, enforcing the same number of iterations for all data
Tm = nm · p · K · T0 , (28)
channels may excessively sift more extrema to the following
where Tm denotes the computation cost, and nm is the number modes, thus causes the phenomenon of over-decomposition.
of sifting times consumed for one MIMF extraction. In contrast (ii) According to ref [9], the condition for equality of the
to MEMD, the p-variate input of FMEMD is projected and number of extrema and zero crossings is not imposed for stop-
then operated univariable using the standard EMD. Therefore, ping of the MEMD based sifting process, as extrema cannot be
the total consumption of FMEMD for a p-variate IMIMF properly defined for multivariate signals. This may also lead
computation only involves to incomplete rationality in the decomposition process.
MIMF 8 7 6 5 4 3 2 1
time, where Tf denotes the time cost, while n̄f represents
Spectrum
K
1
P
an average of the sifting times, namely, n̄f = K ni and
i=1
ni , i = 1, 2, . . . , K represents the respective sifting times for
standard EMD applying on the corresponding projected func- 10
0
10
1
10
2
Zero crossings(log 2)
10
IMFs is similar to that applied for the standard IMF, thus it is
8
concluded that nm = max {ni , i = 1, 2, . . . , K}. To this end,
6
combine Eq. (28) with Eq. (29), a new formulation is given MEMD approximate slope = −0.74087
4
Tm ≥ p · Tf . (30) 2
1 2 3 4 5 6 7 8
MIMF index
This inequality illustrates that running of the proposed
FMEMD is p times or more faster than that of the MEMD. In
Fig. 4. Top: averaged spectra of MIMFs (MIMF1-MIMF8 from right to left)
practice, according to our experiments, the multiple is much obtained by MEMD. Bottom: averaged number of base-2 logarithm of zero
larger than p. crossings plotted versus the MIMF index.
x1 x2 x3 x4 x5 x6
10 10 10 10 10 10
0 0 0 0 0 0
−10 −10 −10 −10 −10 −10
5 5 5 5 5 5
d1
0 0 0 0 0 0
−5 −5 −5 −5 −5 −5
5 5 5 5 5 5
d2
0 0 0 0 0 0
−5 −5 −5 −5 −5 −5
5 5 5 5 5 5
d3
0 0 0 0 0 0
−5 −5 −5 −5 −5 −5
5 5 5 5 5 5
d4
0 0 0 0 0 0
−5 −5 −5 −5 −5 −5
5 5 5 5 5 5
0 0 0 0 0 0
r
−5 −5 −5 −5 −5 −5
0 50 100 0 50 100 0 50 100 0 50 100 0 50 100 0 50 100
Sample
x1 x2 x3 x4 x5 x6
10 10 10 10 10 10
0 0 0 0 0 0
−10 −10 −10 −10 −10 −10
5 5 5 5 5 5
id1
0 0 0 0 0 0
−5 −5 −5 −5 −5 −5
5 5 5 5 5 5
id2
0 0 0 0 0 0
−5 −5 −5 −5 −5 −5
5 5 5 5 5 5
id3
0 0 0 0 0 0
−5 −5 −5 −5 −5 −5
5 5 5 5 5 5
id4
0 0 0 0 0 0
−5 −5 −5 −5 −5 −5
5 5 5 5 5 5
0 0 0 0 0 0
r
−5 −5 −5 −5 −5 −5
0 50 100 0 50 100 0 50 100 0 50 100 0 50 100 0 50 100
Sample
Fig. 6. Decomposition results of the synthesized signal using the proposed FMEMD.
TABLE I
O SCILLATION DETECTION RESULTS OF THE DECOMPOSED IMIMF S .
Variable Tag 21 Tag 22 Tag 23 Tag 24 Tag 25 Tag 26 Tag 27 Tag 28 Tag 29 Tag 30
id6 0.3636 0.5686 0.8246 0.8074 0.7499 0.6354 0.5218 0.6373 0.9437 0.6667
id7 0.6114 2.9346 3.0956 3.0562 2.6970 3.0569 1.7429 1.3659 3.0348 0.7561
IEEE ACCESS 12
Tag21 Tag22 Tag23 Tag24 Tag25 Tag26 Tag27 Tag28 Tag29 Tag30
5
0
−5
5
id1−id5
0
−5
5
id6
0
−5
5
id7
0
−5
5
id8
0
−5
5
0
r
−5
0 1000 0 1000 0 1000 0 1000 0 1000 0 1000 0 1000 0 1000 0 1000 0 1000
Sample
theta (4 ∼ 7Hz) and delta (< 4Hz) bands, which are well Recently, the uses of multivariate extension of EMD [34],
explained by id3 , id4 , id5 and id6 respectively. Based on [35], [36], [37] were proposed for the fusion of multiple out-
the aligned common patterns, the underlying brain activity of-focus and multi-exposure images. It was shown that the bi-
regarding different research purposes [31], [32] can be further variate EMD outperforms wavelet and PCA based approaches,
analyzed. On the other hand, the remaining modes of this de- particularly in retaining edge-based information from different
composition also have their corresponding physical meanings: image modalities. In this section, the proposed FMEMD is
(i) id1 : high frequency components which are not completely adopted to merge two multi-exposure images to a single
cleared by pass-band filter; (ii) id2 : power-line interference in improved image as an illustration.
EEG recordings.
The methodology employed for fusion of both gray-level
The FMEMD method enables features of the signal within
and RGB based colored images is similar to the one used
the same subband from different channels to be categorized in
in [34]. More precisely, given two gray scale images, each
the same mode which allows the neural information in differ-
considered as a matrix and denoted by X and Y, the fusion
ent frequency bands can be studied across different channels.
scheme is operated as follows:
The IMIMFs preserve not only the amplitude and frequency
features but also the phase characteristic of synthesized com- (i) X and Y are first converted into row vectors by con-
ponents. Unlike other advanced signal processing techniques catenating their rows.
(e.g., independent component analysis (ICA) and principal
(ii) Two such vectors are combined to form a bivariate
component analysis (PCA).) utilizing statistical algorithms to
signal. Then the FMEMD method is applied to this data and
extract common information, only the cross-channel common
resulting in M IMIMFs.
components are considered which could ignore the phase
difference among different channels. (iii) These bivariate IMIMFs are separated and reconverted
back to image-matrices, which yield M scale images for each
input image, denoted by Xi and Yi , for i = 1, . . . , M .
C. Image Fusion M M
(iv) {Xi }i=1 and {Yi }i=1 are combined locally, based on
Image fusion is the process of registering and combining coefficients computed from the local variance estimates at each
multiple images from different modalities to improve the spatial point, resulting in a fused image Z as given by
imaging quality, while reducing randomness and redundancy M
X
such that the new image is more suitable for the purpose of Z (m, n) = αi (m, n) Xi (m, n) + βi (m, n) Yi (m, n), (33)
i=1
human visual perception and computer-processing tasks [33].
One of the major requirements for successful image fusion is where αi (m, n) and βi (m, n) are weights determined by
that images from the different modalities have to be aligned location (m, n), on the basis of the relative values between
correctly. Moreover, as real world images are usually nonlinear local and total variance for each mode
and nonstationary, which render the traditional techniques var [Xi (m, n)]
invalid, a fully adaptive approach is required [34]. αi (m, n) = , (34)
var [Xi (m, n)] + var [Yi (m, n)]
IEEE ACCESS 13
F3 F4 P3 P4
40
0
−40 Power spectrum
10
id1
0
−10
0 50 100 150
10
id2
0
−10
0 50 100 150
10
id3
0
−10
0 15 30
10
id4
0
−10
0 15 30
10
id5
0
−10
0 15 30
10
id6
0
−10
0 15 30
10
0 Frequency/ Hz
r
−10
0 500 1000 0 500 1000 0 500 1000 0 500 1000
Sample
Fig. 8. FMEMD of 1 s-long normal EEG of channel F3, F4, P3 and P4 (left) with corresponding power spectra of the decomposed IMIMFs (right).
and
var [Yi (m, n)]
βi (m, n) = , (35)
var [Xi (m, n)] + var [Yi (m, n)]
where var [Xi (m, n)] (var [Yi (m, n)]) represents the local
variance of Xi (Yi ) calculated across a small block around
(m, n). Similarly, the RGB based color image fusion can be
extended by applying the methodology to three channels (red,
green and blue) of a color image separately [34].
The performance of the proposed method is evaluated on
two multi-exposure images that are shown in Fig. 9 and Fig.
10 respectively. The first image is obtained from a work table,
which is severely underexposed. Observed that the small plate
in the right region is nearly hidden, whereas the left part,
showing details of the book, is well exposed. In the second
input image, on the other hand, the details on the book are
covered due to the strong light, but the right area is exposed Fig. 9. Image of a work table which is obtained from underexposed condition.
properly. Fig. 11 shows the result of the fusion obtained from
applying FMEMD based scheme. As desired, the fused image
retains all the detail of the original images and any spurious
The mode-aligned property of FMEMD contributes to the
fusion artifacts are kept into a minimum.
automatic clustering of similar frequency modes as well as
providing both time and frequency information in nonstation-
VII. C ONCLUSION ary time series. Compared with MEMD, operation of the
A novel multivariate extension of the standard EMD, proposed FMEMD is independent of the dimension of the
namely the fast MEMD (FMEMD), has been provided. multivariate signal, and this approach avoids the problem of
Three key concepts including the direction-dependent MIMF over-decomposition when processing on the white noise as
(DMIMF), the direction-independent MIMF (IMIMF) and the a filter bank. Moreover, numerical simulation shows that the
pseudo-IMIMF are first defined through properly extending the FMEMD method presents better decomposition performance
notion of multivariate extrema. Then FMEMD is generalized in processing data with low sampling frequency. Such superi-
from the univariate EMD by solving an overdetermined system orities can facilitate more widespread applications of the mul-
of linear equations. It is shown numerically that the proposed tivariate EMD method, especially for the implementations in
FMEMD, similar to the standard EMD, follows a dyadic filter analyzing long data and signals in which the dyadic filter bank
bank structure (channel-wise) for multivariate white Gaussian decomposition is relevant. Simulations and several real world
noise. cases have verified the effectiveness of proposed method.
IEEE ACCESS 14
A. Proof of Homogeneity
Let us suppose s (t) = S (α, t), s∗ (t) = S (kα, t), and
n n
{Mi }i=0 , {Mi∗ }i=0 denote the second derivatives of s (t) and
s∗ (t) on time instants τ = {t0 , t1 , . . . , tn } respectively.
(i) Calculate the related parameters.
Fig. 10. Image of a work table which is obtained from overexposed condition. hi = ti+1 − ti , (36)
f [ti , ti+1 ] = xi+1hi−xi
(
−xi ) , (37)
f ∗ [ti , ti+1 ] = k(xi+1 hi = kf [ti , ti+1 ]
where i = 0, 1, · · · , n − 1. Accordingly, some spline-related
parameters can be further computed
(
hi−1 ∗
µi = hi−1 +hi =µi
∗ , (38)
λi = 1 − µi = λi
f [ti−1 ,ti ]−f [ti ,ti+1 ]
f [ti−1 , ti , ti+1 ] = ti−1 −ti+1
f [ti−1 ,ti ]−f ∗ [ti ,ti+1 ]
∗ ,
f ∗ [ti−1 , ti , ti+1 ] = =kf [ti−1 , ti , ti+1 ]
ti−1 −ti+1
(39)
di = 6f [ti−1 , ti , ti+1 ]
, (40)
di ∗ = 6f ∗ [ti−1 , ti , ti+1 ] = kdi
where i = 1, 2, · · · , n−1. According to [38], these parameters
satisfy the following equation
Fig. 11. The fused image obtained by employing the proposed FMEMD.
µi Mi−1 + 2Mi + λi Mi+1 = di , (41)
n
from i = 1, 2, · · · , n − 1. Therefore, moments {Mi }i=0 can
The proposed method may be improved from following be obtained by solving the above system of linear equations,
aspects: (i) a more efficient EMD algorithm, (ii) algorithm which form the spline interpolation function as given by
to solve the overdetermined system of linear equations, (iii) 3 3
method for generating a more suitable set of direction vectors (ti+1 − t) (t − ti )
s(t) = Mi + Mi+1
(EMD cannot decompose invalid projection, for example when 6hi 6hi
, (42)
the energy of the highest frequency component is much Mi hi 2 ti+1 − t Mi+1 hi 2 t − ti
smaller/larger than that of the remained components) and (iv) a + [ai − ] + [ai+1 − ]
6 hi 6 hi
more scientific criterion to determine the number of directions.
where t ∈ [ti , ti+1 ] , i = 0, 1, . . . , n − 1. Notice that s∗ (t) and
n
{Mi∗ }i=0 show the similar relationship.
A PPENDIX (ii) Set boundary conditions.
P ROOF OF THE L EMMA 1 Since Eq. (41) has n + 1 unknown numbers but only n − 1
equations, the two ends M0 and Mn should be developed in
Lemma 1. The projection of the cubic spline interpolation
advance. Notice that the not-a-knot end condition [39] requires
curve of a multi-dimensional point set is equivalent to the
the third derivatives of the spline function at the first and
interpolation that is calculated from the projected point set.
last interior break are continuous respectively. The boundary
Proof. Let ∆ := {a = t0 < t1 < · · · < tn = b} be a parti- conditions upon M0 and Mn can therefore be calculated as
tion of the time interval [a, b], α = {a0 , a1 , . . . , an } and
M0 = λ11 (M1 − µ1 M2 )
β = {b0 , b1 , . . . , bn } represent two different point sets that 1 . (43)
Mn = µn−1 (Mn−1 − λn−1 Mn−2 )
are defined on time instants τ = {t0 , t1 , . . . , tn } respectively,
then it can be proved that the univariate cubic spline functions Likewise, M0∗ and Mn∗ meet the similar equations.
IEEE ACCESS 15
n n
(iii) Solving of {Mi }i=0 and {Mi∗ }i=0 . (iv) Solving of the moments.
Combining Eq. (38), (41) and (43), a system of linear Substitute Eq. (51) into the corresponding cubic functions
n
equations for moments {Mi }i=0 can be consisted sα (t), sβ (t), sα+β (t) that are defined by Eq. (42), it is
µ1 µ1
2+ 1− obtained that
λ1 λ1
M1
d1
µ2 2 λ2
M2 d2
[2] D. P. Mandic, N. ur Rehman, Z. Wu, and N. E. Huang, “Empirical mode [25] G. Rilling and P. Flandrin, “On the influence of sampling on the empir-
decomposition-based time-frequency analysis of multivariate signals: ical mode decomposition,” in Acoustics, Speech and Signal Processing,
The power of adaptive data analysis,” IEEE Signal Processing Magazine, 2006. ICASSP 2006 Proceedings. 2006 IEEE International Conference
vol. 30, no. 6, pp. 74–86, 2013. on, vol. 3. IEEE, 2006, pp. III–III.
[3] N. E. Huang, Z. Shen, S. R. Long, M. C. Wu, H. H. Shih, Q. Zheng, N.- [26] Z. Xu, B. Huang, and F. Zhang, “Improvement of empirical mode
C. Yen, C. C. Tung, and H. H. Liu, “The empirical mode decomposition decomposition under low sampling rate,” Signal Processing, vol. 89,
and the hilbert spectrum for nonlinear and non-stationary time series no. 11, pp. 2296–2303, 2009.
analysis,” in Proceedings of the Royal Society of London A: Mathemati- [27] C. Park, D. Looney, N. ur Rehman, A. Ahrabian, and D. P. Mandic,
cal, Physical and Engineering Sciences, vol. 454, no. 1971. The Royal “Classification of motor imagery bci using multivariate empirical mode
Society, 1998, pp. 903–995. decomposition,” IEEE Transactions on Neural Systems and Rehabilita-
[4] J. Fleureau, A. Kachenoura, L. Albera, J.-C. Nunes, and L. Senhadji, tion Engineering, vol. 21, no. 1, pp. 10–22, 2013.
“Multivariate empirical mode decomposition and application to multi- [28] N. F. Thornhill, B. Huang, and H. Zhang, “Detection of multiple
channel filtering,” Signal Processing, vol. 91, no. 12, pp. 2783–2792, oscillations in control loops,” Journal of Process Control, vol. 13, no. 1,
2011. pp. 91–100, 2003.
[5] Y. Lei, J. Lin, Z. He, and M. J. Zuo, “A review on empirical mode [29] X. Lang, Z. Zhang, L. Xie, A. Horch, and H. Su, “Time-frequency
decomposition in fault diagnosis of rotating machinery,” Mechanical analysis of plant-wide oscillations using multivariate intrinsic time-scale
Systems and Signal Processing, vol. 35, no. 1, pp. 108–126, 2013. decomposition,” Industrial & Engineering Chemistry Research, vol. 57,
[6] T. Tanaka and D. P. Mandic, “Complex empirical mode decomposition,” no. 3, pp. 954–966, 2018.
IEEE Signal Processing Letters, vol. 14, no. 2, pp. 101–104, 2007. [30] N. F. Thornhill, J. W. Cox, and M. A. Paulonis, “Diagnosis of plant-wide
[7] G. Rilling, P. Flandrin, P. Gonçalves, and J. M. Lilly, “Bivariate empirical oscillation through data-driven analysis and process understanding,”
mode decomposition,” IEEE Signal Processing Letters, vol. 14, no. 12, Control Engineering Practice, vol. 11, no. 12, pp. 1481–1490, 2003.
pp. 936–939, 2007. [31] T. Zhang, L. Dai, Y. Wang, W. J. Freeman, and G. Li, “Eeg spatiotempo-
[8] N. ur Rehman and D. P. Mandic, “Empirical mode decomposition for ral pattern classification of the stimuli on different fingers,” in Advances
trivariate signals,” IEEE Transactions on Signal Processing, vol. 58, in Cognitive Neurodynamics (IV). Springer, 2015, pp. 147–153.
no. 3, pp. 1059–1068, 2010. [32] G. Li, T. Zhang, X. Chen, C. Shang, and Y. Wang, “Effect of intermittent
[9] ——, “Multivariate empirical mode decomposition,” in Proceedings of hypoxic training on hypoxia tolerance based on brain functional connec-
the Royal Society of London A: Mathematical, Physical and Engineering tivity,” Physiological Measurement, vol. 37, no. 12, p. 2299, 2016.
Sciences. The Royal Society, 2009, p. rspa20090502. [33] H. Li, B. Manjunath, and S. K. Mitra, “Multisensor image fusion using
[10] Y. Lv, R. Yuan, and G. Song, “Multivariate empirical mode decomposi- the wavelet transform,” Graphical models and image processing, vol. 57,
tion and its application to fault diagnosis of rolling bearing,” Mechanical no. 3, pp. 235–245, 1995.
Systems and Signal Processing, vol. 81, pp. 219–234, 2016. [34] N. ur Rehman, D. Looney, T. Rutkowski, and D. Mandic, “Bivariate
[11] X. Lang, D. Zhong, L. Xie, and J. Chen, “Application of improved emd-based image fusion,” in Statistical Signal Processing, 2009. SSP’09.
multivariate empirical mode decomposition to plant-wide oscillations IEEE/SP 15th Workshop on. IEEE, 2009, pp. 57–60.
characterization,” in Advanced Control of Industrial Processes (Ad- [35] D. Looney and D. P. Mandic, “Multiscale image fusion using complex
CONIP), 2017 6th International Symposium on. IEEE, 2017, pp. 601– extensions of emd,” IEEE Transactions on Signal Processing, vol. 57,
606. no. 4, pp. 1626–1630, 2009.
[12] N. ur Rehman and D. P. Mandic, “Filter bank property of multivariate [36] N. ur Rehman, S. Ehsan, S. M. U. Abdullah, M. J. Akhtar, D. P. Mandic,
empirical mode decomposition,” IEEE Transactions on Signal Process- K. D. McDonald-Maier et al., “Multi-scale pixel-based image fusion
ing, vol. 59, no. 5, pp. 2421–2426, 2011. using multivariate empirical mode decomposition,” Sensors, vol. 15,
[13] A. S. Bahri, D. D. Warnana, F. A. M. Santos, B. J. Santosa et al., “Fast, no. 5, pp. 10 923–10 947, 2015.
simultaneous and robust vlf-em data denoising and reconstruction via [37] S. M. U. Abdullah, N. ur Rehman, M. M. Khan, and D. P. Mandic, “A
multivariate empirical mode decomposition,” Computers & Geosciences, multivariate empirical mode decompositionbased approach to pansharp-
vol. 67, pp. 125–138, 2014. ening,” IEEE Transactions on Geoscience and Remote Sensing, vol. 53,
[14] Y. H. Wang, C. H. Yeh, H. W. V. Young, K. Hu, and M. T. Lo, no. 7, pp. 3974–3984, 2015.
“On the computational complexity of the empirical mode decomposition [38] J. Stoer and R. Bulirsch, Introduction to numerical analysis. Springer
algorithm,” Physica A Statistical Mechanics & Its Applications, vol. 400, Science & Business Media, 2013, vol. 12.
no. 2, pp. 159–167, 2014. [39] C. De Boor, “Convergence of cubic spline interpolation with the not-
[15] A. Ahrabian, N. ur Rehman, and D. P. Mandic, “Bivariate empirical a-knot condition.” WISCONSIN UNIV-MADISON MATHEMATICS
mode decomposition for unbalanced real-world signals.” IEEE Signal RESEARCH CENTER, Tech. Rep., 1985.
Processing Letters, vol. 20, no. 3, pp. 245–248, 2013. [40] K. E. Atkinson, An introduction to numerical analysis. John Wiley &
[16] N. ur Rehman, M. W. Safdar, U. ur Rehman, and D. P. Mandic, Sons, 2008.
“Dynamically-sampled bivariate empirical mode decomposition,” IEEE [41] R. Bellman, Introduction to matrix analysis. Siam, 1997, vol. 19.
Signal Processing Letters, vol. 21, no. 7, pp. 857–861, 2014.
[17] N. ur Rehman, K. Naveed, M. Safdar, S. Ehsan, and K. McDonald-
Maier, “Dynamically sampled multivariate empirical mode decomposi-
tion,” Electronics Letters, vol. 51, no. 24, pp. 2049–2051, 2015.
[18] J. Stewart, Calculus. Cengage Learning, 2011.
[19] N. E. Huang, M.-L. C. Wu, S. R. Long, S. S. Shen, W. Qu, P. Gloersen,
and K. L. Fan, “A confidence limit for the empirical mode decomposition
and hilbert spectral analysis,” in Proceedings of the Royal Society of
London A: Mathematical, Physical and Engineering Sciences, vol. 459,
no. 2037. The Royal Society, 2003, pp. 2317–2345.
Xun Lang was born in 1994. He received the
[20] H. Niederreiter, Random number generation and quasi-Monte Carlo
B.S. degree in automation from Zhejiang Univer-
methods. SIAM, 1992.
sity, HangZhou, China, in 2014. He is currently
[21] J. A. Cadzow, “Minimum `1 , `2 and `∞ norm approximate solutions
pursuing the Ph.D. degree with Zhejiang University,
to an overdetermined system of linear equations,” Digital Signal Pro-
Hangzhou, China. He is currently with the Institute
cessing, vol. 12, no. 4, pp. 524–560, 2002.
of Cyber-Systems and Control, Zhejiang University.
[22] P. Flandrin, G. Rilling, and P. Goncalves, “Empirical mode decompo-
His current research interests include, but are not
sition as a filter bank,” IEEE Signal Processing Letters, vol. 11, no. 2,
limited to, process control performance monitoring,
pp. 112–114, 2004.
signal processing, time-frequency analysis.
[23] Z. Wu and N. E. Huang, “A study of the characteristics of white noise
using the empirical mode decomposition method,” in Proceedings of
the Royal Society of London A: Mathematical, Physical and Engineering
Sciences, vol. 460, no. 2046. The Royal Society, 2004, pp. 1597–1611.
[24] P. Flandrin, P. Gonçalves, and G. Rilling, “Emd equivalent filter banks,
from interpretation to applications,” in Hilbert–Huang transform and its
applications. World Scientific, 2014, pp. 99–116.
IEEE ACCESS 17
Qian Zheng was born in 1995. He received a Hongye Su received his Ph.D. degree from Zhejiang
B.S. degree in the Department of Automation from University in 1995. He was appointed as a professor
Harbin Engineering University, Harbin, China, in in Dec. 2000 in the Institute of Advanced Control
2017. He is currently pursuing the Ph.D. degree in Zhejiang University and as the deputy director of
at the Institute of Cyber-Systems and Control in the Institute of Advanced Control from Oct. 1999
Zhejiang University, Hangzhou, China. His current to Aug. 2008. Now, he is the deputy director of the
research interests include fault detection and diag- Institute of Cyber-systems and Control in Zhejiang
nosis, image texture analysis, signal processing. University.
His research interests include Process Control &
Optimization Theory and Application. He is a Senior
Member of IEEE, and his IEEE Membership starts
from 2010. He is the author/coauthor of 100 journal papers, 4 books. Professor
Su has undertaken more than ten national and provincial funds and state
key projects as the principal investigator (PI), including National Outstanding
Youth Science Foundation of China, National Natural Science Foundation
Zhiming Zhang was born in 1994. He received a of China, Teaching and Research Award Program for Outstanding Young
B.S. degree in automation from Zhejiang University, Teachers in Higher Education Institutions of MOE, P.R.C., and projects for
Hangzhou, China, in 2017. He is currently pursuing National High Technology Research and Development Program of China (863
the Ph.D. degree at the Institute of Cyber-Systems Program), etc.
and Control in Zhejiang University. His current
research interests include model predictive control,
autonomous vehicle, process control performance
monitoring.