Download presentation
Presentation is loading. Please wait.
1
An Introduction to HHT for Nonlinear and Nonstationary Time Series Analysis: A Plea for Adaptive Data Analysis Norden E. Huang Research Center for Adaptive Data Analysis National Central University
2
What is data? Data (plural of Datum) [Latin: data – what is given] Information; facts, evidence, records, statistics, etc. from which conclusions can be formed. Information in a form suitable for storing and processing by a computer.
3
Data “ In God we trust ” ; everyone else has to show data. “ 我們相信上帝 ”; 其他人都得有數據.
4
The 21st century is a century dominated by data of all forms. Data, data everywhere!
5
GoogleTrend © : Happy
6
GoogleTrend © : Love love
7
GoogleTrend © : Crisis crisis 1.00 crisis
8
Ever since the advance of computer, there is an explosion of data. The situation has changed from a thirsty for data to that of drinking from a fire hydrant. We are drowning in data, but thirsty for knowledge!
9
Data Processing and Data Analysis Processing [proces < L. Processus < pp of Procedere = Proceed: pro- forward + cedere, to go] : A particular method of doing something. Data ProcessingData Processing >>>> Mathematically meaningful parameters Analysis [Gr. ana, up, throughout + lysis, a loosing] : A separating of any whole into its parts, especially with an examination of the parts to find out their nature, proportion, function, interrelationship etc. Data AnalysisData Analysis >>>> Physical understandings
10
Data Analysis Why we do it? How did we do it? What should we do?
11
Why?
12
Why do we have to analyze data? Data are the only connects we have with the reality; data analysis is the only means we can find the truth and deepen our understanding of the problems.
13
Ever since the advance of computer and sensor technology, there is an explosion of very complicate data. The situation has changed from a thirsty for data to that of drinking from a fire hydrant.
14
Henri Poincaré Science is built up of facts *, as a house is built of stones; but an accumulation of facts is no more a science than a heap of stones is a house. * Here facts are indeed our data.
15
Data and Data Analysis Data Analysis is the key step in converting the ‘facts’ into the edifice of science. It infuses meanings to the cold numbers, and lets data telling their own stories and singing their own songs.
16
Science vs. Philosophy Data and Data Analysis are what separate science from philosophy: With data we are talking about sciences; Without data we can only discuss philosophy.
17
Scientific Activities Collecting, analyzing, synthesizing, and theorizing are the core of scientific activities. Theory without data to prove is just hypothesis. Therefore, data analysis is a key link in this continuous loop.
18
Data Analysis Data analysis is too important to be left to the mathematicians. Why?!
19
Different Paradigms I Mathematics vs. Science/Engineering Mathematicians Absolute proofs Logic consistency Mathematical rigor Scientists/Engineers Agreement with observations Physical meaning Working Approximations
20
Different Paradigms II Mathematics vs. Science/Engineering Mathematicians Idealized Spaces Perfect world in which everything is known Inconsistency in the different spaces and the real world Scientists/Engineers Real Space Real world in which knowledge is incomplete and limited Constancy in the real world within allowable approximation
21
Rigor vs. Reality As far as the laws of mathematics refer to reality, they are not certain; and as far as they are certain, they do not refer to reality. Albert Einstein
22
How?
23
Data Processing vs. Analysis In pursue of mathematic rigor and certainty, however, we lost sight of physics and are forced to idealize, but also deviate from, the reality. As a result, we are forced to live in a pseudo-real world, in which all processes are Linear and Stationary
24
削足適履 Trimming the foot to fit the shoe.
25
Available Data Analysis Methods for Nonstationary (but Linear) time series Spectrogram Wavelet Analysis Wigner-Ville Distributions Empirical Orthogonal Functions aka Singular Spectral Analysis Moving means Successive differentiations
26
Available Data Analysis Methods for Nonlinear (but Stationary and Deterministic) time series Phase space method Delay reconstruction and embedding Poincar é surface of section Self-similarity, attractor geometry & fractals Nonlinear Prediction Lyapunov Exponents for stability
27
Typical Apologia Assuming the process is stationary …. Assuming the process is locally stationary …. As the nonlinearity is weak, we can use perturbation approach …. Though we can assume all we want, but the reality cannot be bent by the assumptions.
28
掩耳盜鈴 Stealing the bell with muffed ears
29
The Real World Mathematics are well and good but nature keeps dragging us around by the nose. Albert Einstein
30
Motivations for alternatives: Problems for Traditional Methods Physical processes are mostly nonstationary Physical Processes are mostly nonlinear Data from observations are invariably too short Physical processes are mostly non-repeatable. ÈEnsemble mean impossible, and temporal mean might not be meaningful for lack of stationarity and ergodicity. ÈTraditional methods are inadequate.
31
What?
32
The job of a scientist is to listen carefully to nature, not to tell nature how to behave. Richard Feynman To listen is to use adaptive methods and let the data sing, and not to force the data to fit preconceived modes. The Job of a Scientist
33
How to define nonlinearity? Based on Linear Algebra: nonlinearity is defined based on input vs. output. But in reality, such an approach is not practical. The alternative is to define nonlinearity based on data characteristics.
34
Characteristics of Data from Nonlinear Processes
35
Duffing Pendulum x
36
Duffing Equation : Data
37
Hilbert Transform : Definition
38
Hilbert Transform Fit
39
Conformation to reality rather then to Mathematics We do not have to apologize, we should use methods that can analyze data generated by nonlinear and nonstationary processes. That means we have to deal with the intrawave frequency modulations, intermittencies, and finite rate of irregular drifts. Any method satisfies this call will have to be adaptive.
40
The Traditional Approach of Hilbert Transform for Data Analysis
41
Traditional Approach a la Hahn (1995) : Data LOD
42
Traditional Approach a la Hahn (1995) : Hilbert
43
Traditional Approach a la Hahn (1995) : Phase Angle
44
Traditional Approach a la Hahn (1995) : Phase Angle Details
45
Traditional Approach a la Hahn (1995) : Frequency
46
Why the traditional approach does not work?
47
Hilbert Transform a cos + b : Data
48
Hilbert Transform a cos + b : Phase Diagram
49
Hilbert Transform a cos + b : Phase Angle Details
50
Hilbert Transform a cos + b : Frequency
51
The Empirical Mode Decomposition Method and Hilbert Spectral Analysis Sifting
52
Empirical Mode Decomposition: Methodology : Test Data
53
Empirical Mode Decomposition: Methodology : data and m1
54
Empirical Mode Decomposition: Methodology : data & h1
55
Empirical Mode Decomposition: Methodology : h1 & m2
56
Empirical Mode Decomposition: Methodology : h3 & m4
57
Empirical Mode Decomposition: Methodology : h4 & m5
58
Empirical Mode Decomposition Sifting : to get one IMF component
59
Two Stoppage Criteria : S and SD A.The S number : S is defined as the consecutive number of siftings, in which the numbers of zero- crossing and extrema are the same for these S siftings. B. SD is small than a pre-set value, where
60
Empirical Mode Decomposition: Methodology : IMF c1
61
Definition of the Intrinsic Mode Function (IMF)
62
Empirical Mode Decomposition Sifting : to get all the IMF components
63
Empirical Mode Decomposition: Methodology : data & r1
64
Empirical Mode Decomposition: Methodology : data and m1
65
Empirical Mode Decomposition: Methodology : data, r1 and m1
66
Empirical Mode Decomposition: Methodology : IMFs
67
Definition of Instantaneous Frequency
68
Definition of Frequency Given the period of a wave as T ; the frequency is defined as
69
Equivalence : The definition of frequency is equivalent to defining velocity as Velocity = Distance / Time
70
Instantaneous Frequency
71
The combination of Hilbert Spectral Analysis and Empirical Mode Decomposition is designated as HHT (HHT vs. FFT)
72
Jean-Baptiste-Joseph Fourier 1807 “On the Propagation of Heat in Solid Bodies” 1812 Grand Prize of Paris Institute “Théorie analytique de la chaleur” ‘... the manner in which the author arrives at these equations is not exempt of difficulties and that his analysis to integrate them still leaves something to be desired on the score of generality and even rigor. ’ 1817 Elected to Académie des Sciences 1822 Appointed as Secretary of Math Section paper published Fourier’s work is a great mathematical poem. Lord Kelvin
73
Comparison between FFT and HHT
74
Comparisons: Fourier, Hilbert & Wavelet
75
Speech Analysis Hello : Data
76
Four comparsions D
77
Fourier analysis is incapable of representing any variation in temporal. It is not even capable to separate noise from delta functions!
78
Noise and Delta Functions Movies
81
An Example of Sifting & Time-Frequency Analysis
82
Length Of Day Data
83
LOD : IMF
84
Orthogonality Check Pair-wise % 0.0003 0.0001 0.0215 0.0117 0.0022 0.0031 0.0026 0.0083 0.0042 0.0369 0.0400 Overall % 0.0452
85
LOD : Data & c12
86
LOD : Data & Sum c11-12
87
LOD : Data & sum c10-12
88
LOD : Data & c9 - 12
89
LOD : Data & c8 - 12
90
LOD : Detailed Data and Sum c8-c12
91
LOD : Data & c7 - 12
92
LOD : Detail Data and Sum IMF c7-c12
93
LOD : Difference Data – sum all IMFs
94
Traditional View a la Hahn (1995) : Hilbert
95
Mean Annual Cycle & Envelope: 9 CEI Cases
96
Mean Hilbert Spectrum : All CEs
97
Tidal Machine
98
Properties of EMD Basis The Adaptive Basis based on and derived from the data by the empirical method satisfy nearly all the traditional requirements for basis a posteriori: Complete Convergent Orthogonal Unique
99
Hilbert ’ s View on Nonlinear Data
100
Duffing Type Wave Data: x = cos(wt+0.3 sin2wt)
101
Duffing Type Wave Perturbation Expansion
102
Duffing Type Wave Wavelet Spectrum
103
Duffing Type Wave Hilbert Spectrum
104
Duffing Type Wave Marginal Spectra
105
Duffing Equation
106
Duffing Equation : Data
107
Duffing Equation : IMFs
108
Duffing Equation : Hilbert Spectrum
109
Duffing Equation : Detailed Hilbert Spectrum
110
Duffing Equation : Wavelet Spectrum
111
Duffing Equation : Hilbert & Wavelet Spectra
112
Speech Analysis Nonlinear and nonstationary data
113
Speech Analysis Hello : Data
114
Four comparsions D
115
Global Temperature Anomaly Annual Data from 1856 to 2003
116
Global Temperature Anomaly 1856 to 2003
117
IMF Mean of 10 Sifts : CC(1000, I)
118
Statistical Significance Test
119
Data and Trend C6
120
Rate of Change Overall Trends : EMD and Linear
121
What This Means Instantaneous Frequency offers a total different view for nonlinear data: instantaneous frequency with no need for harmonics and unlimited by uncertainty. Adaptive basis is indispensable for nonstationary and nonlinear data analysis HHT establishes a new paradigm of data analysis
122
Comparisons FourierWaveletHilbert Basisa priori Adaptive FrequencyIntegral transform: Global Integral transform: Regional Differentiation: Local PresentationEnergy-frequencyEnergy-time- frequency Nonlinearno yes Non-stationarynoyes Uncertaintyyes no Harmonicsyes no
123
Conclusion Adaptive method is the only scientifically meaningful way to analyze data. It is the only way to find out the underlying physical processes; therefore, it is indispensable in scientific research. It is physical, direct, and simple.
124
History of HHT 1998: The Empirical Mode Decomposition Method and the Hilbert Spectrum for Non-stationary Time Series Analysis, Proc. Roy. Soc. London, A454, 903-995. The invention 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 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.
125
Recent Developments in HHT 2007: On the trend, detrending, and variability of nonlinear and nonstationary time series. Proc. Natl. Acad. Sci., 104, 14,889-14,894. The correct adaptive trend determination method 2009: On Ensemble Empirical Mode Decomposition. Advances in Adaptive Data Analysis. Advances in Adaptive data Analysis, 1, 1-41 2009: On instantaneous Frequency. Advances in Adaptive Data Analysis (in press) 2009: Multi-Dimensional Ensemble Empirical Mode Decomposition. Advances in Adaptive Data Analysis (Accepted; Patent Pending)
126
Current Efforts and Applications Non-destructive Evaluation for Structural Health Monitoring –(DOT, NSWC, DFRC/NASA, KSC/NASA Shuttle, THSR) Vibration, speech, and acoustic signal analyses –(FBI, and DARPA) Earthquake Engineering –(DOT) Bio-medical applications –(Harvard, Johns Hopkins, UCSD, NIH, NTU, VHT, AS) Climate changes –(NASA Goddard, NOAA, CCSP) Cosmological Gravity Wave –(NASA Goddard) Financial market data analysis –(NCU) Theoretical foundations –(Princeton University and Caltech)
128
HHT looks wonderfully simple, but The Devil is in the details.
129
The Idea behind EMD To be able to analyze data from the nonstationary and nonlinear processes and reveal their physical meaning, the method has to be Adaptive. Adaptive requires a posteriori (not a priori) basis. But the present established mathematical paradigm is based on a priori basis. Only a posteriori basis could fit the varieties of nonlinear and nonstationary data without resorting to the mathematically necessary (but physically nonsensical) harmonics.
130
The Idea behind EMD The method has to be local. Locality requires differential operation to define properties of a function. Take frequency, for example. The present established mathematical paradigm is based on Integral transform. But integral transform suffers the limitation of the uncertainty principle.
131
John von Neumann By and large it is uniformly true that in mathematics there is a time lapse between a mathematical discovery and the moment it becomes useful; and that this lapse can be anything from 30 to 100 years, in some cases even more; and that the whole system seems to function without any direction, without any reference to usefulness, and without any desire to do things which are useful.
132
Norden E. Huang : paraphrase By and large it is usually true that in science there is a time lapse between the discovery of a useful method and the moment it becomes mathematically proved; and that this lapse can be anything from 30 to 100 years, in some cases even more; and that the whole system seems to function with a firm direction: always reference to usefulness, and with strong desire to do things which are useful.
133
On Calculus “Newton and Leibniz's approach to the calculus fell well short of later standards of rigor. We now see their "proof" as being in truth mostly a heuristic hodgepodge mainly grounded in geometric intuition.” Wikipedia 1643-17271646-1716
134
On Calculus George Berkeley, in a tract called The Analyst and elsewhere, ridiculed this and other aspects of the early calculus, pointing out that natural science grounded in the calculus required just as big of a leap of faith as theology grounded in Christian revelation. Modern, rigorous calculus only emerged in the 19th century, thanks to the efforts of Augustin Louis Cauchy, Bernhard Riemann, Karl Weierstrass, and others, who based their work on the definition of a limit and on a precise understanding of real numbers.
135
Jean-Baptiste-Joseph Fourier (1768-1830) 1807 “On the Propagation of Heat in Solid Bodies” 1812 Grand Prize of Paris Institute “Théorie analytique de la chaleur” ‘... manner in which the author arrives at these equations is not exempt of difficulties and that the his analysis to integrate them still leaves something to be desired on the score of generality and even rigor. ’ 1817 Elected to Académie des Sciences 1822 Appointed as Secretary of Académie paper published Fourier’s work is a great mathematical poem. Lord Kelvin
136
Fourier Transform Michel Plancherel (16 January 1885 to 4 March 1967) : Plancherel, Michel (1910) "Contribution a l'etude de la representation d'une fonction arbitraire par les integrales définies," Rendiconti del Circolo Matematico di Palermo, vol. 30, pages 298-335. In which he proved the convergence of the Fourier transform by excluding denumerable number of discontinuity points in the function. He then proved that the totality of the excluded parts have zero measure.
137
Oliver Heaviside
138
Oliver Heaviside 1850 - 1925 Adapted complex numbers to the study of electrical circuits, invented mathematical techniques to the solution of differential equations (later found to be equivalent to Laplace transform), introduced delta and step functions, and invented modern vector analysis, thereby reducing the original twenty equations in twenty unknowns down to the four differential equations in two unknowns we now know as Maxwell’s Equations. Why should I refuse a good dinner simply because I don't understand the digestive processes involved.
139
I will not wait for the Mathematician’s proof. Historically, the proof would not change the method, but will change the mathematician’s view.
140
John von Neumann As a mathematical discipline travels far from its empirical source, or still more, if it is a second and third generation only indirectly inspired by ideas coming from “reality,” it is beset with very grave dangers. It becomes more and more purely aestheticizing, more and more purely l’art pour l’art. This need not be bad if the field is surrounded by correlated subjects, which still have closer empirical connections, or if the discipline is under the influence of men [or women] with exceptionally well-developed taste. But there is a grave danger that the subject will developed along the line of least resistance, that the stream, so far from its source, will separate into a multitude of insignificant branches, and the discipline will become a disorganized mass of details and complexities.
141
I prefer the Heaviside principle. For the time being.
142
HHT is at the same stage as Wavelet was in the late 1980s. We need someone, as Ingrid Daubechies, to set the rigorous foundation. But, in principle, HHT is much harder!
143
VOLUME I TECHNICAL PROPOSAL AND MANAGEMENT APPROACH Mathematical Analysis of the Empirical Mode Decomposition Ingrid Daubechies 1 and Norden Huang 2 1 Program in Applied and Computational Mathematics (Princeton) 2 Research Center for Adaptive Data Analysis, (National Central University) Since its invention by PI Huang over ten years ago, the Empirical Mode Decomposition (EMD) has been applied to a wide range of applications. The EMD is a two-stage, adaptive method that provides a nonlinear time- frequency analysis that has been remarkably successful in the analysis of nonstationary signals. It has been used in a wide range of fields, including (among many others) biology, geophysics, ocean research, radar and medicine. …….
144
The Battle Hymn for HHT By Ingrid Daubechies You can use wavelets or Fourier And find something that is useful, But if you want something new to say, You cannot be any old fool. You then find a new kind of algorithm And let it loose on the world. Before they even know what hit’em, They’re up to their necks on work. Rushing in to meet your frustrating challenge, They filter and stretch and squeeze. But always some signal throws a monkey wrench, And make them huff and sneeze. Your name is, of course, Norden Huang, Long may you live and smile! We’re here to learn and get the hang And will not quit for a very long while. Hilbert-Huang transform, you will triumph! Composed on 17 December 2008 in Guangzhou, to be sung to the tune of ‘The Internationale’. At The Second International Conference on the Advances of Hilbert-Huang Transform and its Applications.
145
Up Hill Does the road wind up-hill all the way? Yes, to the very end. Will the day’s journey take the whole long day? From morn to night, my friend. --- Christina Georgina Rossetti
146
A less poetic paraphrase There is no doubt that our road will be long and that our climb will be steep. …… But, anything is possible. --- Barack Obama 18 Jan 2009, Lincoln Memorial
147
It is better lucky than smart. Everyone needs luck sometimes: Albert Michelson, Carl Wilson, Arno Penzias and Robert Wilson,…
148
I am lucky to have found this simple method. Now we need smart people to tell us why it works and why it works so well. Good luck to us all!!
149
Outline of the Course I Introduction EMD and EEMD EMD Intermittency and confidence limit EEMD Orthogonality End effects Relationship with Fourier decomposition: a conjecture Trend and detrend HHT Operations
150
Outline of the Course II Hilbert Spectral Analysis Mathematical preliminary End effects Wavelet and Wigner-Ville Distribution Instantaneous Frequency Paradoxes of instantaneous frequency Hilbert Spectral Representation and Marginal spectrum Multi-dimensional EMD Available approaches MD-EEMD Applications Water wave studies HHT based nondestructive Health monitoring Stability spectral analysis Global Climate Change
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.