Asian Transactions on Engineering (ATE ISSN: 2221-4267) Volume 01 Issue 05

Denoising of EMG Signals Based on Wavelet Transform

N. M. SOBAHI
Dept. of Electrical and Computer Engineering
KING ABDULAZIZ UNIVERSITY
JEDDAH, SAUDI ARABIA
nsobahi@ieee.org

Abstract— Wavelet analysis is often very effective because problems of EMG signals analysis and justify the accepted
it provides a simple approach for dealing with local aspects of a measures.
signal. Electromyography (EMG) signals can be used for The technology of EMG recording is relatively new. There
clinical/biomedical applications, Evolvable Hardware Chip (EHW) are still limitations in detection and characterization of existing
development, and modern human computer interaction. EMG
nonlinearities in the surface electromyography (sEMG, a special
signals acquired from muscles require advanced methods for
detection, decomposition, processing, and classification. In this technique for studying muscle signals) signal, estimation of the
paper, Wavelet transform (WT) has been applied for removing phase, acquiring exact information due to derivation from
noise from the surface EMG. To fully understand the concept of normality [1][2] Traditional system reconstruction algorithms
WT, Matlab Simulation was used for sEMG data was collected from have various limitations and considerable computational
a forearm muscle. complexity and many show high variance [1]. Recent advances in
technologies of signal processing and mathematical models have
Keywords: sEMG; wavelet transform; denoising; Daubechies; made it practical to develop advanced EMG detection and
decomposition analysis techniques. Various mathematical techniques and
Artificial Intelligence (AI) have received extensive attraction.
I. INTRODUCTION
Mathematical models include wavelet transform, time-frequency
Biomedical signal means a collective electrical signal approaches, Fourier transform, Wigner-Ville Distribution
acquired from any organ that represents a physical variable of (WVD), statistical measures, and higher-order statistics.
interest. This signal is normally a function of time and is The use of Fourier analysis to study biological signals such
describable in terms of its amplitude, frequency and phase. The as EMG recordings is not the most efficient method for transient
EMG signal is a biomedical signal that measures electrical data analysis. However, the time frequency analysis based on the
currents generated in muscles during its contraction representing wavelet transform is better suited to handle the non-stationary
neuromuscular activities. The nervous system always controls the characteristics of the EMG signals.
muscle activity (contraction/relaxation). Hence, the EMG signal
is a complicated signal, which is controlled by the nervous II. EMG: ANATOMICAL AND PHYSIOLOGICAL
system and is dependent on the anatomical and physiological BACKGROUND
properties of muscles. EMG signal acquires noise while traveling The development of EMG started with Francesco Redi’s
through different tissues. Moreover, the EMG detector, documentation in 1666. The document informs that highly
particularly if it is at the surface of the skin, collects signals from specialized muscle of the electric ray fish generates electricity
different motor units at a time which may generate interaction of [3]. By 1773, Walsh had been able to demonstrate that Eel fish’s
different signals. Detection of EMG signals with powerful and muscle tissue could generate a spark of electricity. In 1792, a
advance methodologies is becoming a very important publication entitled ―De Viribus Electricitatis in Motu Musculari
requirement in biomedical engineering. The main reason for the Commentarius‖ appeared, written by A. Galvani, where the
interest in EMG signal analysis is in clinical diagnosis and author showed that electricity could initiate muscle contractions
biomedical applications. The shapes and firing rates of Motor [4]. Six decades later, in 1849, Dubios-Raymond discovered that
Unit Action Potentials (MUAPs) in EMG signals provide an it was also possible to record electrical activity during a voluntary
important source of information for the diagnosis of muscle contraction. The first recording of this activity was made
neuromuscular disorders. Once appropriate algorithms and by Marey in 1890, who also introduced the term
methods for EMG signal analysis are readily available, the nature electromyography [5]. In 1922, Gasser and Erlanger used an
and characteristics of the signal can be properly understood and oscilloscope to show the electrical signals from muscles. Because
hardware implementations can be made for various EMG signal of the stochastic nature of the myoelectric signal, only rough
related applications. So far, research and extensive efforts have information could be obtained from its observation. The
been made in the area, developing better algorithms, upgrading capability of detecting electromyographic signals improved
existing methodologies, improving detection techniques to steadily from the 1930s through the 1950s and researchers began
reduce noise, and to acquire accurate EMG signals. Few to use improved electrodes more widely for the study of muscles
hardware implementations have been done for prosthetic hand [1]. Clinical use of surface EMG for the treatment of more
control, grasp recognition, and human machine interaction. It is specific disorders began in the 1960s. Hardyck and his
quite important to carry out an investigation to classify the actual researchers were the first (1966) practitioners to use sEMG [5]. In
the early 1980s, Cram and Steger introduced a clinical method for

Nov 2011 ATE-50104056©Asian-Transactions 17

 Asian Transactions on Engineering (ATE ISSN: 2221-4267) Volume 01 Issue 05

scanning a variety of muscles using an EMG sensing device [5]. and skin. Therefore, EMG signal will be vastly greater complex
It is not until the middle of the 1980s that integration techniques signal while traveling through different layers; in consequence of
in electrodes had sufficiently advanced to allow batch production acquiring noise coming from the crosstalk between lots of surface
of the required small and lightweight instrumentation and acquired signals those propagate in the same time in different
amplifiers. At present a number of suitable amplifiers are motor units in different layers.
commercially available. In the early 1980s, cables became
available which produce artifacts in the desired microvolt range.
During the past 15 years, research has resulted in a better
understanding of the properties of surface EMG recording. In
recent years, surface electromyography is increasingly used for
Figure 1. Surface raw EMG signal
recording from superficial muscles in clinical protocols, where
intramuscular electrodes are used for deep muscle only [2][4].
Electromyography (EMG) signals can be employed for III. WAVELET ANALYSIS
clinical and biomedical applications. At present, there are three
A transform can be thought of as a remapping of a signal that
common applications of the EMG signal, first, determining the
provides more information than the original. The Fourier
activation timing of the muscle; that is, when the excitation to the
transform fits this definition quite well because the frequency
muscle begins and ends; second, estimating the force produced
information it provides often leads to new insights about the
by the muscle; third, obtaining an index of the rate at which a
original signal. Fourier analysis provides a good description of
muscle fatigues through the analysis of the frequency spectrum of
the frequencies in a waveform, but not their timing. However, the
the signal.
inability of the Fourier transform to describe both time and
Muscle contraction and other activities in the human body
frequency characteristics of the waveform led to a number of
are processed in the brain. Then, the brain will dictate to spinal
different approaches. None of these approaches was able to
cord. The motor neurons in the spinal cord will divide to a
completely solve the time–frequency problem. Timing
number of axonal which is called axonal terminals. Axonal
information is often of primary interest in many biomedical
terminals are attached to a number of muscle fiber. The electrical
signals. A wide range of approaches have been developed to try
signal propagates in the muscle fibers. Muscle fibers are
to extract both time and frequency information from a waveform.
innervated by neurons whose cell bodies are located in spinal
Basically they can be divided into two groups: time–frequency
cord. The nerve fibers, or axons, of these motor neurons leave the
methods and time–scale methods. The wavelet transform can be
spinal cord and are distributed to the motor nerves. As mentioned
used as yet another way to describe the properties of a waveform
above, each motor axon branches several times and innervates
that changes over time, but in this case the waveform is divided
many muscle fibers. A single motor neuron and all innervated
not into sections of time, but segments of scale [7].
muscle fibers are named a motor unit. The distributed signal in
one motor unit is synchronized and equal for all innervated IV. SHORT-TERM FOURIER TRANSFORM: THE
muscle fibers. The resulting signal is called the muscle fiber SPECTROGRAM
action potential. The combination of the muscle fiber action
The first time–frequency methods were based on the
potentials from all the muscle fibers of a single motor unit is the
straightforward approach of slicing the waveform of interest into
motor unit action potential (MUAP). MUAP in EMG signals
a number of short segments and performing the analysis on each
provide an important source of information for the diagnosis of
of these segments, usually using the standard Fourier transform.
neuromuscular disorders. When a motor unit fires, all of the
A window function is applied to a segment of data, effectively
innervated muscle fibers in a motor unit are fired. When a motor
isolating that segment from the overall waveform, and the Fourier
unit is recurrent firing, the motor unit will create a train of
transform is applied to that segment. This is termed the
impulses known as the motor unit action potential train
spectrogram or ―short-term Fourier transform‖ (STFT) since the
(MUAPT). The combination of electrical activity created by each
Fourier Transform is applied to a segment of data that is shorter,
active motor unit is the myoelectrical signal (ME) [6].
often much shorter, than the overall waveform. Since abbreviated
To acquire surface EMG (sEMG) signal, electrodes are
data segments are used, selecting the most appropriate window
placed on the skin overlying the muscle. Alternatively, wire or
length can be critical. This method has been successfully applied
needle electrodes are used and these can be placed directly in the
in a number of biomedical applications.
muscle. When EMG is acquired from electrodes mounted
There are two main problems with the spectrogram: (1)
directly on the skin, the signal is a composite of all the muscle
selecting an optimal window length for data segments that
fiber action potentials occurring in the muscle or muscles
contain several different features may not be possible, and (2) the
underlying the skin. Hence, the EMG signal is a complicated
time–frequency tradeoff: shortening the data length, N, to
signal, which is controlled by the nervous system and is
improve time resolution will reduce frequency resolution which
dependent on the anatomical and physiological properties of
is approximately 1/(NTs). Shortening the data segment could also
muscles. The EMG signal may be either positive or negative
result in the loss of low frequencies that are no longer fully
voltage as shown in Fig. 1.
included in the data segment. Hence, if the window is made
This premise was necessary to illustrate the EMG signal
smaller to improve the time resolution, then the frequency
how it is produced and also to understand the complexity of this
resolution is degraded and vice versa.
signal whereas, as aforementioned, the recording usually done by
A number of approaches have been developed to overcome
mounting the electrodes on the skin overlying thousands of
some of the shortcomings of the spectrogram. The first of these
muscle fibers. Moreover, there are a variety of layers of
was the Wigner-Ville distribution which is also one of the most
connective tissue (tissue which connects other tissues and organs)

Nov 2011 ATE-50104056©Asian-Transactions 18

 Asian Transactions on Engineering (ATE ISSN: 2221-4267) Volume 01 Issue 05

studied and best understood of the many time–frequency autocorrelation function to mitigate the damage done by the cross
methods. The approach was actually developed by Wigner for products. In addition, the Wigner-Ville distribution can have
use in physics, but later applied to signal processing by Ville, negative regions that have no meaning. The Wigner-Ville
hence the dual name. The Wigner-Ville distribution is a special distribution also has poor noise properties. Essentially the noise is
case of a wide variety of similar transformations known under the distributed across all time and frequency including cross products
heading of Cohen’s class of distributions. The Wigner-Ville of the noise, although in some cases, the cross products and noise
distributions, and others of Cohen’s class, use an approach that influences can be reduced by using a window. In such cases, the
harkens back to the early use of the autocorrelation function for desired window function is applied to the lag dimension of the
calculating the power spectrum. The classic method for instantaneous autocorrelation function (Eq.4).
determining the power spectrum was to take the Fourier In the Fourier transform, the waveform was compared to
transform of the autocorrelation function. To construct the a sine function - in fact, a whole family of sine functions at
autocorrelation function, the waveform is compared with itself harmonically related frequencies. This comparison was carried
for all possible relative shifts, or lags. The equations are written out by multiplying the waveform with the sinusoidal functions,
in both continuous and discreet forms: then averaging (using either integration in the continuous
domain, or summation in the discrete domain):
(1)
and (5)
(2)
where Eq. (5) is in the continuous form.
where  and n are the shift of the waveform with respect to itself. Almost any family of functions could be used to probe
In the standard autocorrelation function, time is the characteristics of a waveform, but sinusoidal functions are
integrated (or summed) out of the result, and this result, rxx(), is particularly popular because of their unique frequency
only a function of the lag, or shift, . The Wigner-Ville, and in characteristics: they contain energy at only one specific
fact all of Cohen’s class of distributions, use a variation of the frequency. Naturally, this feature makes them ideal for probing
autocorrelation function where time remains in the result. This is the frequency makeup of a waveform, i.e., its frequency
achieved by comparing the waveform with itself for all possible spectrum. Other probing functions can be used, functions chosen
lags, but instead of integrating over time, the comparison is done to evaluate some particular behavior or characteristic of the
for all possible values of time. This comparison gives rise to the waveform. If the probing function is of finite duration, it would
defining equation of the so-called instantaneous autocorrelation be appropriate to translate, or slide, the function over the
function: waveform, x(t), as is done in the short-term Fourier transform
(STFT):
(3)
(6)
(4)
where f, the frequency, also serves as an indication of family
member, and w(t-) is some sliding window function where t acts
where  and n are the time lags as in autocorrelation, and *
to translate the window over x. More generally, a translated
represents the complex conjugate of the signal, x.
probing function can be written as:
As mentioned above, the classic method of computing
the power spectrum was to take the Fourier transform of the
standard autocorrelation function. The Wigner-Ville distribution (7)
echoes this approach by taking the Fourier transform of the
instantaneous autocorrelation function, but only along the  (i.e., where f(t)m is some family of functions, with m specifying the
lag) dimension. The result is a function of both frequency and family number.
time. When the one dimensional power spectrum was computed If the family of functions, f(t)m, is sufficiently large, then it
using the autocorrelation function, it was common to filter the should be able to represent all aspects the waveform x(t). This
autocorrelation function before taking the Fourier transform to would then allow x(t) to be reconstructed from X(t,m) making this
improve features of the resulting power spectrum. While no such transform bilateral. Often the family of basis functions is so large
filtering is done in constructing the Wigner-Ville distribution, all that X(t,m) forms a redundant set of descriptions, more than
of the other approaches apply a filter (in this case a sufficient to recover x(t). This redundancy can sometimes be
two-dimensional filter) to the instantaneous autocorrelation useful, serving to reduce noise or acting as a control, but may be
function before taking the Fourier transform. In fact, the primary simply unnecessary. Note that while the Fourier transform is not
difference between many of the distributions in Cohen’s class is redundant, most transforms represented by Eq. (7) (including the
simply the type of filter that is used. STFT and all the distributions discussed before) would be, since
The Wigner-Ville has several advantages over the STFT, but they map a variable of one dimension (t) into a variable of two
also has a number of shortcomings. However, the Wigner-Ville dimensions (t,m).
distribution has a number of shortcomings. Most serious of these
V. THE CONTOUNUS WAVELET TRANSFORM
is the production of cross products: the demonstration of energies
at time–frequency values where they do not exist. These phantom The wavelet transform introduces an intriguing twist to the
energies have been the prime motivator for the development of basic concept defined by Eq. (7). In wavelet analysis, a variety of
other distributions that apply various filters to the instantaneous different probing functions may be used, but the family always

Nov 2011 ATE-50104056©Asian-Transactions 19

 Asian Transactions on Engineering (ATE ISSN: 2221-4267) Volume 01 Issue 05

consists of enlarged or compressed versions of the basic function, analysis applications such as described above, but will be costly
as well as translations. This concept leads to the defining if the application calls for recovery of the original signal. For
equation for the continuous wavelet transform (CWT): recovery, all of the coefficients will be required and the
computational effort could be excessive. In applications that
(8) require bilateral transformations, we would prefer a transform
that produces the minimum number of coefficients required to
recover accurately the original signal. The discrete wavelet
where b acts to translate the function across x(t) just as t does in
transform (DWT) achieves this parsimony by restricting the
the equations above, and the variable a acts to vary the time scale
variation in translation and scale, usually to powers of 2. The
of the probing function, . If a is greater than one, the wavelet basic analytical expressions for the DWT will be presented here;
function, , is stretched along the time axis, and if it is less than however, the transform is easier to understand, and easier to
one (but still positive) it contacts the function. Negative values of implement using filter banks. The DWT is often introduced in
a simply flip the probing function on the time axis. While the terms of its recovery transform:
probing function  could be any of a number of different
functions, it always takes on an oscillatory form, hence the term
―wavelet.‖ The * indicates the operation of complex conjugation, (11)
and the normalizing factor 1/ ensures that the energy is the
same for all values of a (all values of b as well, since translations Here k is related to a as: a = 2k; b is related to l as b = 2k l; and
do not alter wavelet energy). If b = 0, and a = 1, then the wavelet d(k,l) is a sampling of W(a,b) at discrete points k and l.
is in its natural form, which is termed the mother wavelet;* that In the DWT, a new concept is introduced termed the scaling
function, a function that facilitates computation of the DWT. To
is, 1,0(t)  (t). A mother wavelet along with some of its family
implement the DWT efficiently, the finest resolution is computed
members produces by dilation and contraction. The wavelet
first. The computation then proceeds to coarser resolutions, but
shown is the popular Morlet wavelet, named after a pioneer of
rather than start over on the original waveform, the computation
wavelet analysis, and is defined by the equation:
uses a smoothed version of the fine resolution waveform. This
smoothed version is obtained with the help of the scaling
(9) function. In fact, the scaling function is sometimes referred to as
the smoothing function. The definition of the scaling function
uses a dilation or a two-scale difference equation:
The wavelet coefficients, W(a,b), describe the correlation
between the waveform and the wavelet at various translations and
(12)
scales: the similarity between the waveform and the wavelet at a
where c(n) is a series of scalars that defines the specific scaling
given combination of scale and position, a, b.
function. This equation involves two time scales (t and 2t) and
If the wavelet function, ψ(t), is appropriately chosen, then it is
can be quite difficult to solve.
possible to reconstruct the original waveform from the wavelet
In the DWT, the wavelet itself can be defined from the
coefficients just as in the Fourier transform. Since the CWT
scaling function:
decomposes the waveform into coefficients of two variables, a
and b, a double summation (or integration) is required to recover
the original signal from the coefficients: (13)

where d(n) is a series of scalars that are related to the waveform

x(t) (Eq. (11)) and that define the discrete wavelet in terms of the
(10) scaling function. While the DWT can be implemented using the
Where above equations, it is usually implemented using filter bank
techniques.

VII. WAVELET DENOISING

The Surface EMG (sEMG) signals was denoised using
and 0 < C < ∞ (the so-called admissibility condition) for recovery discrete wavelet transform (DWT) and a threshold method. The
using Eq. (10). DWT and threshold based denoising was implemented using
In fact, reconstruction of the original waveform is rarely MATLAB Wavelet toolbox. The figure below shows the flow of
performed using the CWT coefficients because of the redundancy the algorithm.
in the transform. When recovery of the original waveform is
desired, the more parsimonious discrete wavelet transform is Wavelet
Threshold
Wavelet
sEMG Decomposing Reconstruction
used. Methods

VI. THE DISCRETE WAVELET TRANSFORM

The CWT has one serious problem: it is highly redundant (In Figure 2: Wavelet based denoising of sEMG signals
its continuous form, it is actually infinitely redundant). The CWT
Wavelets commonly used for denoising biomedical signals
provides an oversampling of the original waveform: many more
include the Daubechies (db2, db8, and db6) wavelets and
coefficients are generated than are actually needed to uniquely
orthogonal Meyer wavelet. The wavelets are generally chosen
specify the signal. This redundancy is usually not a problem in
whose shapes are similar to those of the MUAP.

Nov 2011 ATE-50104056©Asian-Transactions 20

 Asian Transactions on Engineering (ATE ISSN: 2221-4267) Volume 01 Issue 05

1. Wavelet Decomposition
The WT decomposes a signal into several multi-resolution
components according to a basic function which is wavelet
function. As discussed before, filters are one of the most widely
used signal processing functions. The resolution of the signal,
which is a measure of the amount of detail information in the
signal, is determined by the filtering operations, and the scale is
determined by upsampling and downsampling operations. The
DWT is computed by successive lowpass and highpass filtering
of the discrete time-domain signal.

2. Threshold Method
Suppose the following equation represents a simple model of
the EMG signal,

(13)

where, s(t), n(t) denotes EMG signals and White Gaussian Noise Figure 3: Decomposition at level 4: s=a4+d4+d3+d2+d1 (before applying
N(0, ), respectively. denoising)
The energy of the original signal s(t) is effectively captured,
to a high percentage, by transform values whose magnitude
are all greater than a threshold, Ts >0. The noise signal’s
transform values all have the magnitudes while lie below a noise
threshold Tn satisfy Tn < Ts. Then, the noise in f(t) can be
removed by thresholding its transform. All values of its transform
whose magnitude lies below the noise threshold T n are set equal
to 0.

VIII. SIMULATIONS AND RESULTS

The raw sEMG data was downloaded from [8]. Data in this
file was obtained using a bipolar surface EMG sensor (emgPLUX
sensor from bioPLUX) placed at a forearm muscle. The sampling Figure 4: Show the Wavelet Tree for db6 of level 4
rate and test time are 1000Hz and 5 second, respectively.
Any of the WFs (db2, db6, db8, and dmey) are effective for By using manual Soft Minmax Thresholding Method, the
noise removal in the case of sEMG based on [9, 10]. In this denoised signal was advanced after the residuals signal was
experiment WF db6 is chosen and found to be effective for noise removed. For each level, the threshold was manually selected as
removal. This work was done by using wavemenu (Wavelet depicts in Figs. 5, 6, and 7, whereas un-scaled white noise
Toolbox Main Menu) in Matlab® Software. Different ways to structure was selected.
denoise the sEMG signal was performed as depicted and
illustrated below.
In Fig. 3 illustrates a sample raw sEMG signal and its
coefficients by using db6 with 4 levels of decomposition.
Moreover, wavelet tree for db6 of level 4 is shown in Fig. 4.

Nov 2011 ATE-50104056©Asian-Transactions 21

 Asian Transactions on Engineering (ATE ISSN: 2221-4267) Volume 01 Issue 05
Figure 7: The black signal is the denoised signal and the red one is the original
signal.

In another hand, another way for choosing the threshold

method, the Compression was used to see the electrical
consumption signal is redisplayed in red with the compressed
version superimposed in black as shown in Fig. 8.

Figure 5: Original details coefficients and their threshold levels. Figure 8: The original and compressed signals are colored red and black,
respectively.

You can see that the compression process removed most of

the noise, but preserved 88.28% of the energy of the signal. The
automatic thresholding was very efficient, zeroing out all but
11.76% of the wavelet coefficients. The residuals signal is shown
in Fig. 9.

Figure 9: The residuals signal.

Figure 6: Threshold levels and their coefficients. From level 1 to 4 the
thresholding are 0.022, 0.109, 0.303 and 0.742 respectively.
IX. CONCLUSION
This paper provides a brief introduction of the wavelet
transform in EMG signals processing. Wavelet denoising
methods is expected to offer a powerful compliment to
conventional filtering techniques like notch filters and frequency
domain filtering methods, which will be very efficient for sEMG
signal analysis. The future study will construct feature vectors
that are entered into a BP neural network classifier and nearest
neighbor classifier, respectively, to add completeness to this
ongoing study.

REFERENCES

Nov 2011 ATE-50104056©Asian-Transactions 22

 Asian Transactions on Engineering (ATE ISSN: 2221-4267) Volume 01 Issue 05

[1] Shahid S. Higher Order Statistics Techniques Applied to

EMG Signal Analysis and Characterization. Ph.D.
thesis, University of Limerick; Ireland, 2004.
[2] Nikias CL, Raghuveer MR. Bispectrum estimation: A
digital signal processing framework. IEEE Proceedings
on Communications and Radar 1987; 75(7):869-891.
[3] Basmajian JV, de Luca CJ. Muscles Alive – The
Functions Revealed by Electromyography. The
Williams & Wilkins Company; Baltimore, 1985.
[4] Kleissen RFM, Buurke JH, Harlaar J, Zilvold G.
Electromyography in the biomechanical analysis of
human movement and its clinical application. Gait
Posture 1998; 8(2):143-158.
[5] Cram JR, Kasman GS, Holtz J. Introduction to Surface
Electromyography. Aspen Publishers Inc.;
Gaithersburg, Maryland, 1998.
[6] Reaz MB, Hussain MS, Mohd-Yasin F. Techniques of
EMG signal analysis: detection, processing,
classification and applications. Biol Proced
Online. 2006;8(1):11–35. doi: 10.1251/bpo115.
[7] Semmlow, John. L., "Biosignal and Bimedical Image
Processing MATLAB-BASED APPLICATIONS,"
Rebert Wood Johnson Medical School, New
Brunswick, New Jersey, US.
[8] Retrieved from
"http://www.opensignals.net/index.php?title=Image:E
mgPLUX.txt‖.
[9] T. David Mewette, N. Homer, J. R. Karen,
―Removing Power Line Noise from Recorded EMG‖,
Proceedings of the 23rd annual international conference,
pp 2190- 2193, Oct 25-28, Istanbul, Turkey.
[10] P. W. Mark, ―Wavelet-based noise removal for
biomechanical signals: A comparative study‖, IEEE
Trans. on biomedical engineering, vol 47, no. 2, pp.
360-360.

Nov 2011 ATE-50104056©Asian-Transactions 23