International Journal of Electrical, Electronics and Communication Engineering (IJEEC)

Vol. 1, Issue. 1, April 2015

ISSN (Online): 2395-6747

Empirical Wavelet Transform and its Application

1

1,2

Anjana Francis, 2C.Muruganantham

Department of ECE, Nehru College of Engineering and Research Centre
Thrissur, INDIA
represents different modes. In order to divide the spectrum
into N segments, we need a total of N+1 boundaries, but the
limit of Fourier spectrum is in between 0 and we need a total
of n-1 extra boundaries. To find such boundaries first detect
the local maxima in the spectrum and arrange them in the
decreasing order. [1]
With this maxima plus 0 and we define the boundaries n
of each segment. This is the centre between two consecutive
maxima.(0=0 and n=).

AbstractIn signal processing, adaptive representation of signal

is very important. Empirical Wavelet Transform is a new
adaptive signal decomposition technique. This is very useful for
denoising, decompression etc. This paper presents an adaptive
denoising technique using Empirical wavelet transform.
Experiments presented showing the effectiveness of this method
on one dimensional and two dimensional signals.
Index TermsEmpirical Wavelet Transform, Adaptive, One
Dimension, Two Dimension.

I. INTRODUCTION
In signal processing, time frequency analysis consist
of those techniques that study signal in both time and
frequency domain simultaneously. Signal analysis in
adaptive manner finds a variety of applications in the field
of signal processing. In general we can represent a signal as
a linear combination of basis functions. In wavelet and
Fourier transform these basis functions are derived
independently, but in adaptive techniques these functions
are derived from the information contained in the signal.
Jerome Gilles proposed a new approach to build adaptive
wavelets that is known as Empirical Wavelet Transform.
This method is able to separate the AM-FM components of
the signal. This AM-FM components have a compact
support Fourier spectrum.
In the field of signal processing denoising is still a
challenging problem. So many methods are there to remove
noise content from the signal and to recover the original
signal. Each of these methods has their own advantages and
limitations. Wavelet transform analysis has been widely
used for the purpose of denoising. Traditional denoising
schemes are based on linear methods. That is not suitable
for nonlinear and non -stationary signals. To perform signal
denoising in nonlinear and nonstationary signals an
adaptive signal denoising method using Empirical wavelet
transform is proposed in this paper.

Fig 1 .Partitioning of the Fourier axis

The Empirical wavelets are defined as band pass
filters. The Empirical scaling function and Empirical
wavelet can be expressed as follows.[1]

and,

It can be defined in the same way as that of classical

wavelet transform.
The detail coefficients are given by the inner product with
the Empirical wavelets

II. EMPIRICAL WAVELET TRANSFORM

Empirical wavelets, is a type of wavelet that is adapted to
the processed signal. The construction empirical wavelet is
equivalent to the construction of band pass filters. Each mode
have a compactly supported Fourier spectrum. This spectrum
of modes is centered around a specific frequency.
Assume that the Fourier spectrum is divided into N
segments. There is a limit between each segment i.e, n.
Segmentation of Fourier spectrum provides adaptability. Our
aim is to separate different portions of the spectrum that

And the approximation coefficients by the inner product

with the scaling function

International Journal of Electrical, Electronics and Communication Engineering (IJEEC)

Vol. 1, Issue. 1, April 2015
ISSN (Online): 2395-6747
surrounding pixels. An image containing salt-and-pepper
noise will have dark pixels in bright regions and bright pixels
in dark regions.
Where

n()

Gaussian noise: Each pixel in the image will be changed

from its original value by a small amount. A histogram, a plot
of the amount of distortion of a pixel value against the
frequency with which it occurs, shows a normal distribution of
noise. While other distributions are possible, the Gaussian
(normal) distribution is usually a good model, due to
the central limit theorem that says that the sum of different
noises tends to approach a Gaussian distribution.

and 1() are defined by

Poisson noise: Otherwise known as short noise. This is a

type of electronic noise and can be modeled by a poisson
process. Short noise is originated by the discrete nature of
electric charge.
IV.EXPERIMENTAL RESULTS
For 1D signals, the experiments were performed on an
ECG signal. This noisy ECG signal consists of random noise
and baseline wander. ECG is an example for nonlinear and
nonstationary signal.
For 2D signals experiments were performed on different
images with different types of noise input.PSNR and MSE
value table for each image is tabulated.

The reconstruction is obtained by,

EMPIRICAL WAVELET TRANSFORM ON IMAGES

The idea used for one dimensional Empirical wavelet
transform can be extended to Two dimensional signals also.
Here, the basic idea is that process the rows and columns of
the input image by using 1D EWT . While using images, there
is no chance for two different rows or columns have the same
number of Fourier supports.
If there is different frequency bands, the different frequency
bands mixed together and provide a strange representation. To
overcome this, use the same filters for each rows and columns.

Fig 2. Noisy Input Signal

Fig 3. Cosine tapered filter

Fig 4. Spectrum

III.APPLICATION
A. Denoising
Noise reduction is the process of removing noise from
a signal. De-noising plays a vital role in the field of the signal
processing. It attempts to remove noise and retains the
significant information. The main aim of denoising algorithm
is to reduce the noise level, while preserving the signal
features. Different types of noises are there which affects the
signal quality. Some of them are Salt and pepper noise,
Gaussian noise, Poisson noise.
Salt and pepper noise : Each pixel in the image is very
different in colour or intensity from their surrounding pixels.
The value of a noisy pixel bears no relation to the colour of

Fig 5.Modes extracted after filtering

International Journal of Electrical, Electronics and Communication Engineering (IJEEC)

Vol. 1, Issue. 1, April 2015
ISSN (Online): 2395-6747

ITERATION
NUMBER
1
2
3
4
5

BEFORE
DENOISING
19.8681
16.8578
15.0969
13.8475
12.8784

V.CONCLUSION
Empirical Wavelet transform is a nonlinear and nonstationary signal analysis method. Using this method
nonlinear and non-stationary part of the signal can be
separated. The AM-FM signals have a compactly separated
Fourier spectrum. Separating different modes is equivalent
to segment the Fourier spectrum and the application of some
filtering techniques. Experimental results show the
effectiveness of EWT on 1D and 2D signals. Experiments
are performed using matlab software. For 1D signal analysis
ECG signal is taken as reference. Different images were also
taken for analysis. For analysis Signal to Noise ratio value is
calculated for 1D signals and Peak Signal to Noise Ratio and
Mean Squared Error value for 2D signals.

AFTER
DENOISING
26.8237
25.5278
24.3058
23.1805
22.2878

Table 1. SNR Values for Different iterations

REFERENCES
[1] Jerome Gilles Empirical Wavelet Transform IEEE
Transactions on Signal Processing, Vol. 61, No. 16, August 15,
2013 pp.3999-4010.
[2] N. E. Huang, Z. Shen, S. R. Long, M. C. Wu, H. H. Shih, Q.
Zheng, N.-C. Yen, C. C. Tung, and H.H.Liu, The empirical
mode decomposition and the Hilbert spectrum for nonlinear and
non-stationary time series analysis, Proc. Roy. Soc. London A,
vol. 454, pp. 903995, 1998.
[3] Ramesh Kumar, Prabhat Patel Signal Denoising with Interval
Dependent Thresholding Using DWT and SWT International
Journal of Innovative Technology and Exploring Engineering
(IJITEE) ISSN: 2278-3075, Volume-I, Issue-6, November 2012.
[4] P. Auscher and G.L. Weiss and M.V. Wickerhauser, Local Sine
and Cosine Bases of Coifman and Meyer and the Construction
of Smooth Wavelets, WaveletsA Tutorial in Theory and
Applications, Academic Press, pp. 237256, 1992.
[5] S. Mallat, A Wavelet Tour of Signal Processing - A sparse way,
Third Edition Elsevier, Academic Press, 2009.
[6] I. Daubechies, Ten Lectures on Wavelets, Society for Industrial
and Applied Mathematics, Cbms-Nsf Regional Conference
Series in Applied Mathematics, 1992.
[7] I. Daubechies and J. Lu and H-T. Wu, Synchrosqueezed wavelet
trans- forms: An empirical mode decomposition-like tool,
Journal of Applied and Computational Harmonic Analysis, vol.
30, No. 2, pp. 243261, 2011.
[8] P. Flandrin and P. Goncalves and G. Rilling, EMD Equivalent
Filter Banks, from Interpretation to Applications, Chapter 3 in
Hilbert-Huang Transform and Its Applications, pp. 5774,
World Scientic, 2005.
[9] P. Auscher and G.L. Weiss and M.V. Wickerhauser, Local Sine
and Cosine Bases of Coifman and Meyer and the Construction
of Smooth Wavelets, WaveletsA Tutorial in Theory and
Applications, Academic Press, pp. 237256, 1992.
[10] P. Flandrin and G. Rilling and P. Goncalves, Empirical mode
decompo- sition as a lter bank, IEEE Signal Processing Letters,
vol. 11, No. 2, pp. 112114, 2004.

Fig 6.Noisy images with red, green, blue component and

denoised image

Fig 7. PSNR Values for different Noise Types

Fig 9.MSE Values for different Noise Types