GP503 - Lecture 16

GP 503: Geophysical Signal Processing
Lecture - 16
E. Chandrasekhar
Department of Earth Sciences
Indian Institute of Technology Bombay
E-mail: esekhar@iitb.ac.in
It is important to note the following basic difference between
Fourier transform and wavelet transform.
In Fourier transform of a nonlinear signal, the Fourier coefficients

of the transformed signal represent the contribution of respective
sine and cosine function at each frequency, as the sine and cosine
functions are the basis functions for Fourier transform.
Whereas, in case of wavelet transform, the mother wavelet

function itself is the basis function, which is compressed and
dilated, while analyzing the signal at different time locations of the
signal, thereby facilitating to extract the presence of high and low
frequencies in the signal respectively, with good resolution.
For example, to identify the sharp discontinuities or spikes or
singularities present in the signal, the wavelet function at
smaller scales (i.e., in the compressed form) could be used.
To identify the low frequency features present in the signal,

larger scales of the same wavelet function (i.e., in the dilated
form) could be used.
Since the mother wavelet is the basis function in wavelet

transform and since many wavelets can be defined and used
to analyze a variety of nonlinear signals, the wavelet
transform has many basis functions.
Wavelet Analysis: Basics
A wavelet is a small wave having finite length in time (in technical
terms known as compactly supported) and represented as a real or
complex-valued function, say, ψ(t), satisfying the following properties
that
(i) it integrates to zero (i.e., having zero average)
∞
∫ψ (t )dt = 0
−∞
(1) Admissibility condition
(ii) is square integrable (i.e., having finite energy)

∞
∫ψ (t )dt = 1
2 (2) Regularity condition
−∞
Equation (1) implies that the function ψ(t) must be oscillatory (so that
the sum of its positive and negative excursions cancel out in
averaging) and equation (2) implies that ψ(t) must be finite in length.
Some wavelets and their shapes
Wavelet Analysis: Basics
The wavelet function, ψ(t), signifying the time-frequency localization
is defined as
1  t −τ 
ψ τ , s (t ) = ψ  (3)
s  s 
where S > 0 indicates the scale (or dilation parameter) and τ
indicates the translation parameter. S is analogous to frequency,
in the sense that larger scales (low frequencies) provide
overall information of the signal and smaller scales (high
frequencies) provide detailed information of the signal. The
τ
translation parameter, , is linked to the location of the window
as it is slided over the signal and thus apparently refers to time
information in the transformed domain.
If two functions f(t) and g(t) are square integrable in ℜ
(i.e., f (t ), g (t ) ∈ L (ℜ) ) then their inner product is
2
given by
〈 f (t ), g (t )〉 = ∫ f (t ) ⋅ g (t )dt* (4)
ℜ
According to (4), the CWT can be defined as the inner product

of the mother wavelet ψ(t) and the signal, f (t), given by
1  t −τ 
CWTt (τ , s ) =
s
∫ f (t ) ⋅ψ 
 s 
dt (5)
τ Translation parameter s Scale parameter
f (t) Signal ψ (t ) Wavelet function (mother wavelet)

Equation (5) explains that the wavelet transformation gives a
measure of the similarity between the signal and the wavelet
function.
Such a measure at any particular scale s0 and translation τ0, is

identified by a wavelet coefficient. The larger the value of this
coefficient, the higher the similarity between the signal and the
wavelet at (τ0, s0) and vice-versa.
If more number of high wavelet coefficients occur by using a

particular wavelet, then that indicates higher degree of suitability
of that wavelet to study the signal under investigation.
Two types of Wavelet Transforms
1. Continuous Wavelet Transform (CWT)
2. Discrete Wavelet Transform (DWT)

The CWT Operation
Scale= S1
WT coefficients = C1,1 C2,1 C3,1 C4,1 C5,1 ………

The CWT Operation
Scale= S2

The CWT Operation
Scale= S3

Like-wise, all the CWT coefficients computed at different dilations
and translations are finally expressed in the form of scalograms,
depicting the time-scale representations of the signal. Thus CWT
can easily help identify the space-time localization of singularities,
abrupt jumps, or transients present in the signal.
Because of this, the wavelets are known as ‘mathematical microscopes’.

Some properties of wavelets
1. Symmetry property
The symmetry property wavelets explains that the
wavelet transform (WT) of the mirror (m) of a signal is
mirror of the wavelet transform of the signal.
Mathematically it is explained as
WT{𝑚𝑚 𝑓𝑓 𝑡𝑡 = 𝑚𝑚 WT(𝑓𝑓 𝑡𝑡 )
(Adhikari et al., 2009)

2. Translational invariance property:
In CWT, an amount of small time shift (δt) in the wavelet

function results in the same amount of shift in the transformed
domain. This is known as translational invariance property of
wavelets.
Let f δ t (t ) = f (t − δ t ) be the translation of f (t) by a small time

shift δt. The CWT of f δ t (t )
∞
1  t −τ 
CWTf δ t (τ , s ) = ∫
s −∞
f (t − δ t )ψ 
 s
 dt

∞
1  t ′ − (τ − δ t ) 
= ∫
s −∞
f (t ′)ψ 
 s
 dt

(t ′ = t − δ t )
= CWTf (τ − δ t , s )
Some properties of wavelets
3. Scaling property:
If 𝑓𝑓(𝑥𝑥) has a continuous wavelet transform, given by 𝐶𝐶𝐶𝐶𝐶𝐶𝑓𝑓 (𝑎𝑎, 𝑏𝑏)
1 𝑥𝑥
then the function, 𝑔𝑔(𝑥𝑥), defined as 𝑔𝑔 𝑥𝑥 = 𝑓𝑓 will have the
𝑠𝑠 𝑠𝑠
wavelet transform, given by
𝑎𝑎 𝑏𝑏
𝐶𝐶𝐶𝐶𝐶𝐶𝑔𝑔 𝑎𝑎, 𝑏𝑏 = 𝐶𝐶𝐶𝐶𝐶𝐶𝑓𝑓 ,
𝑠𝑠 𝑠𝑠
Both the shifting and the scaling properties are simple results of
changing variables in the integration.
Scale-Frequency relation
Fc
Fa =
a.D
Fa = Pseudo frequency corresponding to the scale
Fc = Central frequency of a wavelet
a = Scale
D = Sampling interval
Gaus1 Gaus3
Scale Frequency Period Scale Frequency Period
2 0.6667 1.5 2 1.3333 0.75
4 0.3333 3 4 0.6667 1.5
8 0.1667 6 8 0.3333 3
16 0.0833 12 16 0.1667 6
32 0.0417 24 32 0.0833 12
64 0.0208 48 64 0.0417 24
Morlet Sym2
Scale Frequency Period Scale Frequency Period
2 2.7083 0.3692 2 2.2222 0.45
4 1.3542 0.7385 4 1.1111 0.9
8 0.6771 1.4769 8 0.5556 1.8
16 0.3385 2.9538 16 0.2778 3.6
32 0.1693 5.9077 32 0.1389 7.2
64 0.0846 11.8154 64 0.0694 14.4

 Scale of the wavelet has inverse relationship with
the frequency
 The results of the CWT analysis are difficult to

interpret in terms of scales, as they do not have
direct relation with the signal, unlike frequencies
 So, it is important to covert the scales into the

frequencies for better interpretation of signals
 Such a conversion helps to study the frequency

content of the signals under investigation.
EDGE EFFECT REMOVAL:
CWT scalogram showing maximum wavelet coefficient
values at its ends signifies that the data are affected by
‘edge effects’.
Edge effects simply ‘mask’ important features present in

the data. They need to be removed from the data for
better interpretation.
They generally arise due to abrupt cutting of some
portions of the data of interest for use
How to remove them?

SYMMETRIC HALF-POINT METHOD
scale
Sample Number
scale
Chandrasekhar and Rao (2012)

Applications of wavelet analysis in
Geosciences
Wavelet Applications in Geosciences
• Geophysical fluid dynamics (Farge, 1992; Farge et al., 1996)
• Geomagnetism (Alexandrescu et al., 1996; Kunagu et al.,
2013; Chandrasekhar et al., 2013; 2015)
• Electromagnetic induction studies (Zhang and Paulson, 1997;
Zhang et al., 1997; Garcia and Jones, 2008).
• Geophysical well logging
• inter-well relationship (Jansen and Kelkar, 1997)
• determining the sedimentary cycles (Prokoph and
Agterberg, 2000)
• reservoir characterization (Panda et al., 2000; Vega, 2003)
• identifying the depths to the tops of reservoir zones
(Chandrasekhar and Rao, 2012 and references therein; Hill
and Uvarova, 2018, Zhang et al., 2018).
• Seismology and seismic exploration
• Material science and Engineering (Gururajan et al., 2013)
Any Questions Please?

GP503 - Lecture 16

Uploaded by

Copyright:

Available Formats

GP503 - Lecture 16

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

GP503 - Lecture 16

Uploaded by

Copyright:

Available Formats

GP 503: Geophysical Signal Processing

In Fourier transform of a nonlinear signal, the Fourier coefficients

Whereas, in case of wavelet transform, the mother wavelet

To identify the low frequency features present in the signal,

Since the mother wavelet is the basis function in wavelet

(ii) is square integrable (i.e., having finite energy)

According to (4), the CWT can be defined as the inner product

τ Translation parameter s Scale parameter

f (t) Signal ψ (t ) Wavelet function (mother wavelet)

Such a measure at any particular scale s0 and translation τ0, is

If more number of high wavelet coefficients occur by using a

1. Continuous Wavelet Transform (CWT)

2. Discrete Wavelet Transform (DWT)

WT coefficients = C1,1 C2,1 C3,1 C4,1 C5,1 ………

WT coefficients = C1,2 C2,2 C3,2 C4,2 C5,2 ………

WT coefficients = C1,3 C2,3 C3,3 C4,3 C5,3 ………

Because of this, the wavelets are known as ‘mathematical microscopes’.

(Adhikari et al., 2009)

In CWT, an amount of small time shift (δt) in the wavelet

Let f δ t (t ) = f (t − δ t ) be the translation of f (t) by a small time

Scale Frequency Period Scale Frequency Period

2 0.6667 1.5 2 1.3333 0.75

4 0.3333 3 4 0.6667 1.5

Scale Frequency Period Scale Frequency Period

2 2.7083 0.3692 2 2.2222 0.45

4 1.3542 0.7385 4 1.1111 0.9

8 0.6771 1.4769 8 0.5556 1.8

16 0.3385 2.9538 16 0.2778 3.6

32 0.1693 5.9077 32 0.1389 7.2

64 0.0846 11.8154 64 0.0694 14.4

 The results of the CWT analysis are difficult to

 So, it is important to covert the scales into the

 Such a conversion helps to study the frequency

Edge effects simply ‘mask’ important features present in

How to remove them?

Chandrasekhar and Rao (2012)

You might also like