Nonstationary and

Nonlinear
Time Series Analysis
using the Hilbert-Huang
Transform
Norden E. Huang
Goddard Institute for Data Analysis
NASA Goddard Space Flight Center

7/21/2004 1

Outline
Introduction

The Empirical Mode Decomposition (EMD) method, sifting

Intrinsic Mode Function (IMF) components, the adaptive basis through EMD

Confidence limit, degree of stationarity, and statistical significance of IMF

A different view on nonlinearity

Applications and examples

Limitations of HHT and unfinished work

Contact information

7/21/2004 2

Nonstationary and Nonlinear Time Analysis 1

 Intro: Motivations
Physical processes are mostly nonstationary

Physical processes are mostly nonlinear

Data from observations are invariably too short

Physical processes are mostly nonrepeatable

∪ Ensemble mean impossible, and temporal mean

might not be meaningful for lack of ergodicity.
Traditional methods are inadequate.
7/21/2004 3

Intro: Available Data Analysis Methods

for Nonstationary (but Linear) Time Series
Various probability distributions
Spectral analysis and spectrogram
Wavelet analysis
Wigner-Ville distributions
Empirical orthogonal functions (aka singular spectral
analysis)
Moving means
Successive differentiations

7/21/2004 4

Nonstationary and Nonlinear Time Analysis 2

 Intro: Available Data Analysis Methods for
Nonlinear (but Stationary and Deterministic)
Time Series

Phase space method

• Delay reconstruction and embedding
• Poincaré surface of section
• Self-similarity, attractor geometry & fractals
Nonlinear prediction
Lyapunov exponents for stability

7/21/2004 5

Intro: Consequences of these Methods

With the explosion of data and computer,

the field is ready for a data analysis
methodology revolution.

We not only need new methods but also a

new paradigm for analyzing data from
nonlinear and nonstationary processes.

7/21/2004 6

Nonstationary and Nonlinear Time Analysis 3

 Intro: History of EMD
1998: The Empirical Mode Decomposition Method and the Hilbert Spectrum for
Non-stationary Time Series Analysis, Proc. Roy. Soc. London, A454, 903-995.
The introduction of the basic method of EMD and Hilbert transform for
determining the instantaneous frequency and energy.
1999: A New View of Nonlinear Water Waves – The Hilbert Spectrum, Ann. Rev.
Fluid Mech. 31, 417-457.
Introduction of the intermittence in EMD decomposition.
2003: A confidence Limit for the Empirical mode decomposition and the Hilbert
spectral analysis, Proc. of Roy. Soc. London, A459, 2317-2345.
Establishment of a confidence limit without the ergodic assumption.
2004: A Study of the Characteristics of White Noise Using the Empirical Mode
Decomposition Method, Proc. Roy. Soc. London, (in press)
Defined statistical significance and predictability for IMF from EMD.
2004: On the Instantaneous Frequency, Proc. Roy. Soc. London, (Under review)
Removal of the limitations posted by Bedrosian and Nuttall theorems for
Instantaneous Frequency computations.
7/21/2004 7

Intro: Characteristics of Data from

Nonlinear Processes
d2 x
+ x + ε x 3 = γ cos ω t
dt 2

d2 x
⇒
dt 2
+ x (1 + ε x )2
= γ cos ω t

⇒ Spring w ith position dependent cons tan t ,

int ra − w ave frequency m od ulation ;
therefore , w e need ins tan tan eous frequenc y .

7/21/2004 8

Nonstationary and Nonlinear Time Analysis 4

 Intro: Duffing Pendulum

7/21/2004 9

Intro: Definition of Hilbert Transform

For any x( t ) ∈ L p ,

1 x( τ )
y( t ) =
π
℘ ∫τ t −τ
dτ ,

then , x( t ) and y( t ) are com plex conjugate :

z( t ) = x( t ) + i y( t ) = a( t ) e i θ ( t ) ,

where
y( t )
(
a( t ) = x 2 + y 2 ) 1/ 2
and θ ( t ) = tan − 1
x( t )
.

7/21/2004 10

Nonstationary and Nonlinear Time Analysis 5

 Intro: Hilbert Transform Fit

7/21/2004 11

Intro: Traditional View a la Hahn (Length of Day

Data, 1995)

7/21/2004 12

Nonstationary and Nonlinear Time Analysis 6

 Intro: Traditional View a la Hahn (Hilbert, 1995)

7/21/2004 13

Intro: Traditional View a la Hahn (Phase Angle,

1995)

7/21/2004 14

Nonstationary and Nonlinear Time Analysis 7

 Intro: Traditional View a la Hahn (Phase Angle,
1995)

7/21/2004 15

Intro: Traditional View a la Hahn (Frequency,

1995)

7/21/2004 16

Nonstationary and Nonlinear Time Analysis 8

 Why doesn’t the traditional
approach work?

7/21/2004 17

Intro: Why the traditional view doesn’t

work… Hilbert Transform a cos θ + b (Data)

7/21/2004 18

Nonstationary and Nonlinear Time Analysis 9

 Intro: Why the traditional view doesn’t work…
Hilbert Transform a cos θ + b (Phase Diagram)

7/21/2004 19

Intro: Why the traditional view doesn’t work…

Hilbert Transform a cos θ + b (Phase Angle Details)

7/21/2004 20

Nonstationary and Nonlinear Time Analysis 10

 Intro: Why the traditional view doesn’t work…
Hilbert Transform a cos θ + b (Frequency)

7/21/2004 21

Outline
Introduction

The Empirical Mode Decomposition (EMD) method, sifting

Intrinsic Mode Function (IMF) components, the adaptive basis through EMD

Confidence limit, degree of stationarity, and statistical significance of IMF

A different view on nonlinearity

Applications and examples

Limitations of HHT and unfinished work

Contact information

7/21/2004 22

Nonstationary and Nonlinear Time Analysis 11

 EMD & Sifting: Test Data

7/21/2004 23

EMD & Sifting: Test Data and Mean M1

7/21/2004 24

Nonstationary and Nonlinear Time Analysis 12

 EMD & Sifting: Test Data and H1

7/21/2004 25

EMD & Sifting: Test Data, H1, Mean M2

7/21/2004 26

Nonstationary and Nonlinear Time Analysis 13

 EMD & Sifting: Test Data, H2, Mean M3

7/21/2004 27

EMD & Sifting: Test Data, H4, Mean M5

7/21/2004 28

Nonstationary and Nonlinear Time Analysis 14

 EMD & Sifting: Getting one IMF
Component
Sifting : to get one IMF component
x( t ) − m 1 = h1 ,
h1 − m 2 = h2 ,
.....
.....
hk −1 − m k = hk .

⇒ hk = c 1 .

7/21/2004 29

EMD & Sifting: Two Stoppage

Criteria (S and SD)
A. The S number : S is defined as the
consecutive number of siftings in which the
number of zero-crossing and extrema are the
same for these S siftings.
B. SD is small than a pre-set value, where
2
T hk − 1 ( t ) − hk ( t )
SD = ∑
t =0 hk − 1 2 ( t )
.

7/21/2004 30

Nonstationary and Nonlinear Time Analysis 15

 EMD & Sifting: IMF C1

7/21/2004 31

EMD & Sifting: Definition of the Intrinsic

Mode Function
Any function having the sam e num bers of
zero − cros sin gs and extrem a ,and also having
sym m etric envelopes defined by local m ax im a
and m in im a respectively is defined as an
Intrinsic Mode Function ( IMF ).

All IMF enjoys good Hilbert Transfo rm :

⇒⇒ c( t ) = a( t ) e i θ ( t )

7/21/2004 32

Nonstationary and Nonlinear Time Analysis 16

 EMD & Sifting: Getting all IMF
Components
Sifting : to get all the IMF components
x( t ) − c1 = r1 ,
r1 − c2 = r2 ,
. . .
rn− 1 − cn = rn .

n
⇒ x( t ) − ∑c
j =1
j = rn .

7/21/2004 33

EMD & Sifting: Test Data and Residue R1

7/21/2004 34

Nonstationary and Nonlinear Time Analysis 17

 EMD & Sifting: Definition of Instantaneous
Frequency
The F ourier Transform of the Instrinsic M ode
F unnction , c( t ), gives

W (ω ) = ∫ a( t ) e
i (θ −ω t )
dt
t

B y Stationary phase approxim ation w e have

dθ ( t )
=ω ,
dt

This is defined as the Ins tan tan eous F requency .

7/21/2004 35

EMD & Sifting: Comparison between

FFT and HHT
1. F F T :

∑
iω jt
x( t ) = ℜ aje .
j

2. H H T :

i ∫ ω j(τ )dτ
x( t ) = ℜ ∑ j
a j( t ) e t
.

7/21/2004 36

Nonstationary and Nonlinear Time Analysis 18

 EMD & Sifting: Comparisons Between
Fourier, Hilbert, and Wavelet

7/21/2004 37

Outline
Introduction

The Empirical Mode Decomposition (EMD) method, sifting

Intrinsic Mode Function (IMF) components, the adaptive basis through EMD

Confidence limit, degree of stationarity, and statistical significance of IMF

A different view on nonlinearity

Applications and examples

Limitations of HHT and unfinished work

Contact information

7/21/2004 38

Nonstationary and Nonlinear Time Analysis 19

 IMF Components: Adaptive Basis
Generated by EMD

* Orthogonality †
* Completeness
* Uniqueness
* Convergence

These comprise the traditional check list.

7/21/2004 39

IMF Components: Length Of Day Data

7/21/2004 40

Nonstationary and Nonlinear Time Analysis 20

 IMF Components: LOD IMFs

7/21/2004 41

IMF Components: Orthogonality Check

Pair-wise % Overall %

0.0003 0.0452
0.0001
0.0215
0.0117
0.0022
0.0031
0.0026
0.0083
0.0042
0.0369
0.0400

7/21/2004 42

Nonstationary and Nonlinear Time Analysis 21

 IMF Components: Data & Various Partial Sums

7/21/2004 43

IMF Components: Detailed Length of Day Data and

Sum c8-c12

7/21/2004 44

Nonstationary and Nonlinear Time Analysis 22

 IMF Components: Detail LOD Data and Sum IMF
c7-c12

7/21/2004 45

IMF Components: Difference Between LOD

Data and Sum of All IMFs

7/21/2004 46

Nonstationary and Nonlinear Time Analysis 23

 IMF Components: EMD Generated
Adaptive Basis
Completeness
Given by definition
Convergence
Simple reduced cases can be proven
Orthogonality
Reynolds type decomposition: mean ⊥ fluctuation;
not necessary for nonlinear cases
Uniqueness
With respect to adjustable parameters

7/21/2004 47

Outline
Introduction

The Empirical Mode Decomposition (EMD) method, sifting

Intrinsic Mode Function (IMF) components, the adaptive basis through EMD

Confidence limit, degree of stationarity, and statistical significance of IMF

A different view on nonlinearity

Applications and examples

Limitations of HHT and unfinished work

Contact information

7/21/2004 48

Nonstationary and Nonlinear Time Analysis 24

 Confidence Limit: Confidence Limit for
Fourier Spectrum
The confidence limit for Fourier spectral analysis is
based on ergodic assumption.
It is derived by dividing the data into M sections and
substituting the temporal (or spatial) average as the
ensemble average.
This approach is valid for linear and stationary
processes, and the sub-sections have to be
statistically independent.

7/21/2004 49

Confidence Limit: Confidence Limit for

Fourier Spectrum

Confidence Limit from 7 sections, each 2048 points.

7/21/2004 50

Nonstationary and Nonlinear Time Analysis 25

 Confidence Limit: Confidence Limit for
Hilbert Spectrum
Any data can be decomposed into infinitely
many different component sets.
EMD is a method to generate infinitely many
different IMF representations based on different
sifting parameters.
Some of the IMFs are better than others based
on various properties (e.g., Orthogonal Index).
A confidence limit for Hilbert spectral analysis
can be based on an ensemble of “valid” IMFs
resulting from different sifting parameters S
covering the parameter space fairly.
It is valid for nonlinear and nonstationary
processes.
7/21/2004 51

Confidence Limit: Critical Parameters for

EMD
N: the maximum number of siftings allowed to
extract an IMF.

S: the stoppage criterion, or criterion for

accepting a sifting component as an IMF.

Therefore, the nomenclature for the IMFs is as

follows:
CE(N, S) : for extrema sifting
CC(N, S) : for curvature sifting

7/21/2004 52

Nonstationary and Nonlinear Time Analysis 26

 Confidence Limit: Effects of EMD (Sifting)
To separate data into components of
similar scale
To eliminate ridding waves
To make the results symmetric with
respect to the x-axis and to make the
amplitude more even

Note: The first two are necessary for a valid

IMF, the last effect actually caused the IMF
to lose its intrinsic properties.

7/21/2004 53

Confidence Limit: Orthogonal Index as

Function of N and S Contour

7/21/2004 54

Nonstationary and Nonlinear Time Analysis 27

 Confidence Limit: Orthogonality Index as
Function of N and S

7/21/2004 55

Confidence Limit: IMF CE(100, 2)

7/21/2004 56

Nonstationary and Nonlinear Time Analysis 28

 Confidence Limit: IMF CE(100, 10)

7/21/2004 57

Confidence Limit: Mean Hilbert Spectrum

with All CEs

7/21/2004 58

Nonstationary and Nonlinear Time Analysis 29

 Confidence Limit: Mean and STD of
Marginal Hilbert Spectra

7/21/2004 59

Confidence Limit: Mean Envelope from 11

Different Siftings for LOD Data

7/21/2004 60

Nonstationary and Nonlinear Time Analysis 30

 Confidence Limit: Mean Envelopes for
Annual Cycle IMFs

7/21/2004 61

Degree of Staionarity: Defining the

Degree of Stationarity
Traditionally, stationarity is taken for granted;
it is given; it is an article of faith.
All the definitions of stationarity are too
restrictive.
All definitions of stationarity are qualitative.
A good definition must be quantitative to give
a Degree of Stationarity.

7/21/2004 62

Nonstationary and Nonlinear Time Analysis 31

 Degree of Stationarity: Definition of
Strict Stationarity
For a random var iable x( t ), if

2
〈 x( t ) 〉 p ∞ , 〈 x( t ) 〉 = m, and that

[ x( t1 ), x( t2 ), ... x( tn )] and [ x( t1 + τ ), x( t2 + τ ),... x( tn + τ )]

have the same joi nt distribution for all τ .

7/21/2004 63

Degree of Stationarity: Definition of

Wide Sense Stationarity
For any random var iable x( t ), if

2
〈 x( t ) 〉 p ∞ , 〈 x( t ) 〉 = m, and that

[ x( t1 ), x( t2 )] and [ x( t1 + τ ), x( t2 + τ )]
have the same joi nt distribution for all τ .

Therefore, 〈 x( t1 ) ⋅ x( t 2 ) 〉 = C( t1 − t 2 ) .

7/21/2004 64

Nonstationary and Nonlinear Time Analysis 32

 Degree of Stationarity: Definition of
Statistical Stationarity
Applies if the stationarity definitions are satisfied with
certain degree of averaging.

All averaging involves a time scale. The definition of

this time scale is problematic.

7/21/2004 65

Degree of Stationarity: For a Time-

Frequency Distribution
For a time − frequencydistribution, H( ω ,t ),

1
T ∫t
n( ω ) H( ω ,t ) dt ;

2
1
T
 H( ω ,t )
DS( ω )
T ∫
0
1 − n( ω )  dt .
 

7/21/2004 66

Nonstationary and Nonlinear Time Analysis 33

 Degree of Stationarity: Degree of
Statistical Stationarity for a Time-
Frequency Distribution
For a time − frequency distribution, H ( ω ,t ),

1
n( ω )
T ∫ H ( ω ,t ) dt
t
;

2
1
T
 〈 H ( ω ,t )〉 ∆t 
DS( ω , ∆ t )
T ∫
0
1 −
 n( ω ) 
dt .

7/21/2004 67

Statistical Significance: Methodology

Method is based on observations from Monte

Carlo numerical experiments on 1 million
white noise data points.
All IMFs are generated by 10 siftings.
Fourier spectra are based on 200 realizations
of 4,000 data point sections.
Probability densities are based on 50,000
data point data sections.

7/21/2004 68

Nonstationary and Nonlinear Time Analysis 34

 Statistical Significance: IMF Period
Statistics
1 2 3 4 5 6 7 8 9
IMF
347042 168176 83456 41632 20877 10471 5290 2658 1348
Number
of peaks
Mean 2.881 5.946 11.98 24.02 47.90 95.50 189.0 376.2 741.8

period
Periods in 0.240 0.496 0.998 2.000 3.992 7.958 15.75 31.35 61.75

a year

7/21/2004 69

Statistical Significance: Fourier Spectra of

IMFs F o u rie r S p e c tra o f IM F s
(1 0 **-3 )

1 .5

1
s p e c tru m

0 .5

0
0 1 2 3 4 5 6 7 8 9

S h ifte d F o u rie r S p e c tra o f IM F s

(1 0 **-3 )

0 .8

0 .6
s p e c tru m

0 .4

0 .2

0
1 1 .5 2 2 .5 3 3 .5
7/21/2004 ln T 70

Nonstationary and Nonlinear Time Analysis 35

 Statistical Significance: Empirical
Observations I
Normalized spectral area is constant

∫S ln T , n d ln T = const

7/21/2004 71

Statistical Significance: Empirical

Observations II
Computation of mean period

dT d ln T ∫S ln T , n d ln T
NE n = ∫ S ω ,n d ω = ∫ S T ,n T2
= ∫ S ln T , n T
=
Tn

Tn =
∫S ln T , n d ln T
d ln T
∫S ln T , n
T
7/21/2004 72

Nonstationary and Nonlinear Time Analysis 36

 Statistical Significance: Empirical
Observations III
The product of the mean energy and period is
constant

E n T n = const

ln E n + ln T n = const

7/21/2004 73

Statistical Significance: Monte Carlo

Result (IMF Energy vs. Period)

7/21/2004 74

Nonstationary and Nonlinear Time Analysis 37

 Statistical Significance: Empirical Observation,
IMF Histograms
By Central Limit theory IMF should be normally distributed.
5000 m ode 2 5000 m ode 3

0 0
-1 0 1 -1 -0 .5 0 0 .5 1
5000 m ode 4 5000 m ode 5

0 0
-0 .5 0 0 .5 -0 .5 0 0 .5
5000 m ode 6 5000 m ode 7

0 0
-0 .4 -0 .2 0 0 .2 0 .4 -0 .2 0 0 .2
5000 m ode 8 5000 m ode 9

0 0
-0 .2 -0 .1 0 0 .1 0 .2 -0 .1 0 0 .1
7/21/2004 75

Statistical Significance: IMF Energy

Density Histograms
200 200
mode 2 mode 3

100 100

0 0
0.15 0.2 0.25 0.05 0.1 0.15
200 200
mode 4 mode 5

100 100

0 0
0.02 0.04 0.06 0.08 0.01 0.02 0.03 0.04 0.05
200 200
mode 6 mode 7

100 100

0 0
0 0.01 0.02 0.03 0 0.01 0.02
200 300
mode 8 mode 9
200
100
100

0 0
0 0.005 0.01 0 0.005 0.01

By Central Limit Theory, the IMFs should be normally distributed;

therefore, the energy density should be Chi-squared distributed.
7/21/2004 76

Nonstationary and Nonlinear Time Analysis 38

 Statistical Significance: Chi-Squared
Energy Density Distributions

By Central Limit Theory, the IMFs should be normally

distributed; therefore, the energy density should be
Chi-squared distributed.

ρ (E n ) = N ⋅ ( NE n ) NEn 2 −1
e − NE n 2

7/21/2004 77

Statistical Significance: Formula for Confidence

Limit for IMF Distributions
Introduce new variable y:

y = ln E Then, E = ey

 NE 
ρ ( y ) = C ⋅ exp− 1 − y +
( y − y)
2
+
( y − y)
3
+
 
 L
 2  2! 3!  

7/21/2004 78

Nonstationary and Nonlinear Time Analysis 39

 Statistical Significance: Confidence Limit
for IMF Distributions

7/21/2004 79

Statistical Significance: Data and IMFs SOI

S O I

5
C 1 R aw

0
-5
2
0
-2
2
C 2

0
-2
2
C 3

0
-2
2
C 4

0
-2
1
C 5

0
-1
1
C 6

0
-1
1
C 7

0
-1
0 .5
C 8

0
-0 .5
0 .5
C 9

0
-0 .5
0 .2
0
-0 .2
R

-0 .4
1 97/21/2004
30 1940 1950 1960 1970 1980 1990 802 0 0 0

Nonstationary and Nonlinear Time Analysis 40

 Statistical Signifiance: Statistical
Significance for SOI IMFs
IMFs 4, 5, 6 and 7 are 99% statistical significance signals.

1 mon 1 yr 10 yr 100 yr
7/21/2004 81

Statistical Significance: Summary

Not all IMFs have the same statistical significance.
Based on the white noise study, we have established
a method to determine the statistical significant
components.
References:
Wu, Zhaohua and N. E. Huang, 2003: A Study of the
Characteristics of White Noise Using the Empirical Mode
Decomposition Method, Proceedings of the Royal Society of
London (in press).
Flandrin, P., G. Rilling, and P. Gonçalvès, 2003: Empirical
Mode Decomposition as a Filterbank, IEEE Signal
Processing, (in press).

7/21/2004 82

Nonstationary and Nonlinear Time Analysis 41

 Outline
Introduction

The Empirical Mode Decomposition (EMD) method, sifting

Intrinsic Mode Function (IMF) components, the adaptive basis through EMD

Confidence limit, degree of stationarity, and statistical significance of IMFs

A different view on nonlinearity

Applications and examples

Limitations of HHT and unfinished work

Contact information

7/21/2004 83

Nonlinearity: Duffing Type Wave (Data: x =

cos(wt+0.3 sin2wt))

7/21/2004 84

Nonstationary and Nonlinear Time Analysis 42

 Nonlinearity: Duffing Type Wave
(Perturbation Expansion)

For ε 1 , we can have

x( t ) = cos (ω t + ε sin 2ω t )
= cos ω t cos ( ε sin 2ω t ) − sin ω t sin ( ε sin 2ω t )
= cos ω t − ε sin ω t sin 2ω t + ....
 ε ε
=  1 −  cos ω t + cos 3ω t + ....
 2 2

This is very similar to the solutionof Duffing equation .

7/21/2004 85

Nonlinearity: Duffing Type Wave (Wavelet

Spectrum)

7/21/2004 86

Nonstationary and Nonlinear Time Analysis 43

 Nonlinearity: Duffing Type Wave (Hilbert
Spectrum)

7/21/2004 87

Nonlinearity: Duffing Type Wave

(Marginal Spectra)

7/21/2004 88

Nonstationary and Nonlinear Time Analysis 44

 Nonlinearity: Duffing Equation
d2 x
2
+ x + ε x 3 = γ cos ω t .
dt

S o lv e d w ith o d e 2 3 tb fo r t = 0 to 2 0 0 w ith
ε = −1
γ = 0 .1
ω = 0 .0 4 H z

I n itia l c o n d itio n :
[ x ( o ) , x '( 0 ) ] = [ 1 , 1 ]

7/21/2004 89

Nonlinearity: Duffing Equation (Data)

7/21/2004 90

Nonstationary and Nonlinear Time Analysis 45

 Nonlinearity: Duffing Equation (IMFs)

7/21/2004 91

Nonlinearity: Duffing Equation (IMFs)

7/21/2004 92

Nonstationary and Nonlinear Time Analysis 46

 Nonlinearity: Duffing Equation (Hilbert
Spectrum)

7/21/2004 93

Nonlinearity: Duffing Equation (Detailed

Hilbert Spectrum)

7/21/2004 94

Nonstationary and Nonlinear Time Analysis 47

 Nonlinearity: Duffing Equation
(Wavelet Spectrum)

7/21/2004 95

Nonlinearity: Duffing Equation (Hilbert &

Wavelet Spectra)

7/21/2004 96

Nonstationary and Nonlinear Time Analysis 48

 Nonlinearity: Duffing Equation (Marginal
Hilbert Spectrum)

7/21/2004 97

Nonlinearity: Rössler Equation

R&&ossler E q u atio n so lved w ith od e 2 3 :

&x = − ( y + z ),
1
&y = x + y,
5
1
&z = + z( x − µ ).
5

In itital con d ition s :

µ = 3 .5
[ x , y , z ] = [ 1, − 1 , 0 ]

F or
t = 0 : 20 0 .
7/21/2004 98

Nonstationary and Nonlinear Time Analysis 49

 Nonlinearity: Rössler Equation (Data)

7/21/2004 99

Nonlinearity: Rössler Equation (3D

Phase)

7/21/2004 100

Nonstationary and Nonlinear Time Analysis 50

 Nonlinearity: Rössler Equation (2D Phase)

7/21/2004 101

Nonlinearity: Rössler Equation (IMF Strips)

7/21/2004 102

Nonstationary and Nonlinear Time Analysis 51

 Nonlinearity: Rössler Equation (IMFs)

7/21/2004 103

Nonlinearity: Rössler Equation (Hilbert Spectrum)

7/21/2004 104

Nonstationary and Nonlinear Time Analysis 52

 Nonlinearity: Rössler Equation (Hilbert
Spectrum & Data Details)

7/21/2004 105

Nonlinearity: Rössler Equation (Wavelet

Spectrum)

7/21/2004 106

Nonstationary and Nonlinear Time Analysis 53

 Nonlinearity: Rössler Equation (Hilbert &
Wavelet Spectra)

7/21/2004 107

Nonlinearity: Rössler Equation (Marginal

Spectra)

7/21/2004 108

Nonstationary and Nonlinear Time Analysis 54

 Nonlinearity: Rössler Equation (Marginal
Spectra)

7/21/2004 109

Nonlinearity:
What does this mean?
Instantaneous Frequency offers a total different view
for nonlinear data.

An adaptive basis is indispensable for nonstationary

and nonlinear data analysis.

HHT establishes a new paradigm for data analysis.

7/21/2004 110

Nonstationary and Nonlinear Time Analysis 55

 Nonlinearity: Comparisons
Fourier Wavelet Hilbert
Basis A priori A priori Adaptive
Frequency Convolution: Convolution: Differentiation:
Global Regional Local
Presentation Energy- Energy-time- Energy-time-
frequency frequency frequency
Nonlinear No No Yes
Non-stationary No Yes Yes
Feature No Discrete : no Yes
extraction Continuous: yes

7/21/2004 111

Nonlinearity: Different Paradigms

Mathematics vs. Science/Engineering

Mathematicians Scientists/Engineers

Absolute proof Agreement with observations

Logic consistency Physical meaning

Mathematical rigor Working approximations

7/21/2004 112

Nonstationary and Nonlinear Time Analysis 56

 Outline
Introduction

The Empirical Mode Decomposition (EMD) method, sifting

Intrinsic Mode Function (IMF) components, the adaptive basis through EMD

Confidence limit, degree of stationarity, and statistical significance of IMF

A different view on nonlinearity

Applications and examples

Limitations of HHT and unfinished work

Contact information

7/21/2004 113

Applications: Current Applications

Non-destructive evaluation for health monitoring

(DOT, NSWC, and DRC/NASA, KSC Shuttle)
Vibration, speech, and acoustic signal analyses
(FBI, MIT, and DARPA)
Earthquake engineering
(DOT)
Biomedical applications
(Harvard, UCSD, Johns Hopkins, and
Southampton, UK)

7/21/2004 114

Nonstationary and Nonlinear Time Analysis 57

 Applications: Current Applications
Global primary productivity evolution time
series from LandSat data
(NASA Goddard)
Planet hunting
(NASA Goddard and Nicholas Copernicus
University, Poland)
Financial market data analysis
(NASA and HKUST)

7/21/2004 115

Examples: Airfoil Flutter Study

The new NASA aeroelastic flight program is pushing
the airfoil to a new frontier. HHT clearly identified
the yield of the airfoil just before the final
disintegration of the airfoil.

Fourier totally missed the critical change.

7/21/2004 116

Nonstationary and Nonlinear Time Analysis 58

 Examples: Location of the Test Wing

7/21/2004 117

Examples: Details of the Test Wing

7/21/2004 118

Nonstationary and Nonlinear Time Analysis 59

 Examples: Airfoil Flutter

7/21/2004 119

Examples: Full Data

7/21/2004 120

Nonstationary and Nonlinear Time Analysis 60

 Examples: Mean Hilbert Spectrum y(i)

7/21/2004 121

Examples: Mean Hilbert and Spectrogram

y83

7/21/2004 122

Nonstationary and Nonlinear Time Analysis 61

 Examples: Instantaneous Frequency and
Data Envelope

7/21/2004 123

Outline
Introduction

The Empirical Mode Decomposition (EMD) method, sifting

Intrinsic Mode Function (IMF) components, the adaptive basis through EMD

Confidence limit, degree of stationarity, and statistical significance of IMF

A different view on nonlinearity

Applications and examples

Limitations of HHT and unfinished work

Contact information

7/21/2004 124

Nonstationary and Nonlinear Time Analysis 62

 Limitations: Limitations of Hilbert
Transform
• Data need to be mono-component. Traditional
applications using band-pass filter, which distorts the
wave form. (EMD Resolves this problem)

• Bedrosian Theorem: Hilbert transform of [a(t)

cosω(t)] might not be exactly [a(t) sinω(t)] for
arbitrary a and ω . (Normalized HHT resolves this)

• Nuttall Theorem: Hilbert transform of cosω(t) might

not be exactly sinω(t) for arbitrary ω(t). (Normalized
HHT improves on the error bound).

7/21/2004 125

Unfinished Work: Outstanding

Mathematical Problems
1.Adaptive data analysis methodology in general
2.Nonlinear system identification methods
3.Prediction problem for nonstationary processes
(end effects)
4.Optimization problem (the best IMF selection
and the issue of uniqueness, i.e. “Is there a unique
solution?”)
5.Spline problem (best spline implementation of HHT,
convergence, and 2-D)
6.Approximation problem (Hilbert transform
and quadrature)

7/21/2004 126

Nonstationary and Nonlinear Time Analysis 63

 Contact Information
If you are interested in learning more about NASA
Goddard’s HHT technology, please visit our
Website:

http://techtransfer.gsfc.nasa.gov/HHT/HHT.htm

7/21/2004 127

Nonstationary and Nonlinear Time Analysis 64