Professional Documents
Culture Documents
Department of Computer
Science and Electronics
08
A report submitted to the Department of Computer Science and Electronics, Mlardalen University,
in part fulfilment
fulfilment of the Degree of Master of Science in
Electronics with Biomedical Engineering
By
Sheeraz Gul Tareen
Stn05002@student.mdh.se
Tutor:
Mikael Ekstrm
Supervisor:
Prof. Rashid Baig
www.mdh.se
MSc Thesis Report, Mlardalen University, Sweden
Removal of Power Line Interference and other Single Frequency Tones from Signals
Table of Contents
DEDICATION ___________________________________________________________ 7
ACKNOWLEDGEMENTS___________________________________________________ 8
ABSTRACT _____________________________________________________________ 9
1
CHAPTER-1 ________________________________________________________ 10
1.1
1.2
1.3
1.3.1
1.3.2
1.3.3
1.3.4
1.3.5
1.3.6
1.3.7
1.3.8
1.3.9
11
11
11
11
11
12
12
12
12
CHAPTER-2 ________________________________________________________ 13
2.1
2.1.1
Chapter-1 ________________________________________________________________
Chapter-2 ________________________________________________________________
Chapter-3 ________________________________________________________________
Chapter-4 ________________________________________________________________
Chapter-5 ________________________________________________________________
Chapter-6 ________________________________________________________________
Chapter-7 ________________________________________________________________
Chapter-8 ________________________________________________________________
References _______________________________________________________________
2.2
2.3
2.4
2.5
CHAPTER-3 ________________________________________________________ 18
3.1
3.2
3.3
3.4
3.5
3.6
Removal of Power Line Interference and other Single Frequency Tones from Signals
3.7
3.7.1
3.7.2
3.7.3
CHAPTER-4 ________________________________________________________ 24
4.1
Introduction__________________________________________________________ 24
4.2
4.3
4.4
4.5
4.6
4.7
4.8
CHAPTER-5 ________________________________________________________ 31
5.1
Introduction__________________________________________________________ 31
5.2
5.3
5.4
5.4.1
5.4.2
Advantages: ______________________________________________________________ 33
Disadvantages:____________________________________________________________ 33
5.5
5.6
CHAPTER-6 ________________________________________________________ 36
6.1
6.2
6.3
6.4
6.5
Noise Cancellation_____________________________________________________ 39
6.6
6.7
CHAPTER-7 ________________________________________________________ 44
7.1
7.2
Removal of Power Line Interference and other Single Frequency Tones from Signals
7.3
7.4
7.4.1
7.4.2
7.5
CHAPTER-8 ________________________________________________________ 55
8.1
8.1.1
8.1.2
Removal of Power Line Interference (50 Hz) from ECG Signal by LMS Algorithm _______ 46
Removing of Harmonics and High Frequency Noise from Original ECG Signal __________ 51
Conclusion _______________________________________________________________ 55
Suggestion for Future Work/Future Enhancement _______________________________ 56
APPENDIX A _______________________________________________________ 57
9.1
10 APPENDIX C _______________________________________________________ 61
10.1
MATLAB CODE FOR HARMONICS AND HIGH FREQUENCY NOISES FROM ECG SIGNAL
BY GENERAL NOTCH REJECTION FILTER _________________________________________ 61
11 REFERENCES _______________________________________________________ 65
Removal of Power Line Interference and other Single Frequency Tones from Signals
List of Figures
FILTER STRUCTURE
____________________ 34
Removal of Power Line Interference and other Single Frequency Tones from Signals
Removal of Power Line Interference and other Single Frequency Tones from Signals
DEDICATION
Removal of Power Line Interference and other Single Frequency Tones from Signals
ACKNOWLEDGEMENTS
I start with the name of Almighty Allah for providing me with the
unique opportunity to finalise this thesis.
It would not have been possible for me to complete this thesis
Removal of Power Line Interference and other Single Frequency
Tones from Signals,
except for the able guidance and constructive suggestions of my
thesis supervisor (Prof. Rashid Baig), thesis tutors (Mikael
Ekstrm). Their supervision and encouragement has made it
possible for me to complete this task, which is very important for
practical work.
Removal of Power Line Interference and other Single Frequency Tones from Signals
ABSTRACT
Removal of Power Line Interference and other Single Frequency Tones from
Signals
With the latest advancements in electronics, several techniques are used for removal of
unwanted entities from signals especially that are implied in the most sophisticated
applications. The removal of power line interference from most sensitive medical
monitoring equipments can also be removed by implementing various useful
techniques. The power line interference (50/60 Hz) is the main source of noise in most
of bio-electric signals. The thesis report presents the removal of power line interference
and other single frequency tones from ECG signal using the advanced adaptive filtering
technique with LMS (least mean square) algorithm. The thesis is based on digital signal
processing (DSP) techniques with MATLAB package, with the emphases on design of
adaptive LMS algorithm. The adaptive interference removal technique can be used for
removal of power line interference in various potential applications such as recording
Electrocardiograms (ECG), Electroencephalogram (EEG) and Electromyogram (EMG).
MATLAB package will be used in the thesis work which is a powerful tool for the
interactive design in most of the scientific applications and complex engineering
calculations.
As an additional in order to achieve the goal of thesis it will also be investigated and
implemented for the removal of harmonics (hum) and high frequency noise from ECG
signal by using general notch rejection filters, which are partial milestone for the thesis.
Removal of Power Line Interference and other Single Frequency Tones from Signals
1 CHAPTER-1
THESIS INTRODUCTION
1.1 Problem Description
The medical monitoring devices are more sensitive for the biomedical signal recording
and need more accurate results for every diagnosis. It is complicated to get accurate
result for every biomedical signals recording while patient is diagnosis by medical
monitoring equipments such as ECG, EEG and EMG.
The low frequency signal is destroyed by power line interference of 50/60 Hz noise, this
noise is also source of interference for biomedical signal recording. The signal can also
be corrupted by electromagnetic field (EMF) by the machinery which is placed nearby.
[1] The frequency of power line interference 50/60 Hz is nearly equal to the frequency
of ECG, so this 50/60 Hz noise can destroyed the output of ECG signal while the patient
is diagnosis at hospital or some where else. The recording of ECG signal can not give
accurate result due to the power supply or by environment. [1]
There are many reasons for the corruption of ECG signal while recording in hospital or
some other place due to the external interference which comes from power
transformer or high voltage electric power lines and internal interference comes from
the internal power supplies. Other problem occurs by harmonics and high frequency
noises. In a noise signal, the signal component holds harmonics with different amplitude
and frequency. The harmonics frequency is integral multiple of fundamental frequency
such as 50Hz. Due to these interferences the quality of ECG signal can not be ideal so it
is needed to improve the quality of required output of ECG signal.
10
Removal of Power Line Interference and other Single Frequency Tones from Signals
11
Removal of Power Line Interference and other Single Frequency Tones from Signals
1.3.6 Chapter-6
This chapter presents the brief description of adaptive filter algorithm and the least
mean square (LMS) algorithm which is to be employed in this thesis to perform the
noise cancellation. It also gives the overview of the structure & operation of the LMS
algorithm, the design of adaptive filter algorithm, noise cancellation adaptive noise
cancelation and interference cancellation.
1.3.7 Chapter-7
In this chapter the design, analysis and simulation results are described. It also give the
details of main objective & how design by DSP technique and the technique of software
implementation & verification for refinement for signal by LMS, which includes removal
of power line interference from ECG signal & removal of power line interference from
voice signal and also described its simulation results. It also gives the detail and graphs
of removing of humming and high frequency noise from ECG signal. Finally the algorithm
implementation and verification has been presented.
1.3.8 Chapter-8
This chapter provided the conclusion and future research. It also gives the detail of the
thesis goal, its achievement and what has been concluded after completion this thesis.
1.3.9 References
The list of information gathered from books, university library database, journals and
internet sites.
12
Removal of Power Line Interference and other Single Frequency Tones from Signals
2 CHAPTER-2
INTRODUCTION TO ECG SIGNAL WITH POWER LINE INTERFERENCE
2.1 General Overview
The electric power lines are main source of electricity transportation from grid station to
the consumers. Power transformers are used for the transform of voltage which is
generated from the grid station.
The purpose of electricity generation is for the powering electronic & electrical
technologies acquired from various sources of energy. The source of energy for first
power plant was wood, while today it relies mainly on coal, nuclear energy, natural gas,
hydroelectric and petroleum geothermal sources. [2]
Due to the large amount of power involved, transmission normally takes place at high
voltages (110 kV or above). Electricity is usually transmitted over long distance through
overhead power transmission lines. Most transmission lines operate with three-phase
alternating current (AC). The standard frequency in North America is 60 Hz; while 50 Hz
in rest of world.
The power line interference 50/60 Hz is the source of interference in bio potential
measurement and it corrupt the biomedical signals recordings such as
Electrocardiogram (ECG), the Electroencephalogram (EEG) and the Electromyography
(EMG) which are extremely important for the diagnosis of patients. It is hard to find out
the problem because the frequency range of ECG signal is nearly same as the frequency
of power line interference.
The figure 2.1 shows the one period of uncorrupted ECG signal with QRS complex. The
ECG signal contains the information within the frequency range of around 50 Hz that is
why it is called QRS complex. The QRS complex is a waveform which is most important
in all ECGs waveforms and it comes into view in usual and unusual signals in an ECG. [3]
13
Removal of Power Line Interference and other Single Frequency Tones from Signals
Adaptive filters are used to eliminate the power line interference (60 Hz) and they are
proposed to obtain the impulse response of the normal QRS complex. [4] In the figure
above, an uncorrupted ECG signal shows an original signal graph for ECG signal which
demonstrate the diagnosis of heart activities for heart patient.
Consequently, it is analysis that how to remove the power line interference of 50/60 Hz
which is a problem for biomedical signal measurement. Electromagnetic interface (EMI)
from 50/60 Hz power line noise is present in cable holding ECG signal. [5]
Several solutions for the removal of power line interference have been expressed. The
main source of interference is AC power line interference. The interference is caused by
magnetic fields as well as by the electric fields. When special signal recording techniques
are applied, which minimize the interference therefore some AC noise remains as a
consequence of unbalanced input impedances. Further removal of AC noise must be
accomplished either by analog or digital filters.
2.1.1 Heart Mechanism and Purpose of ECG Diagnosis
The heart is a muscular organ, it pump the blood throughout the body and collecting
blood circulating back from the body. [6] Electrical impulses are the main source of
generation of regular normal heartbeat. The heart muscle must be activated electrically
before the beginning of its mechanical function. When the electrical abnormalities of
the heart occur then the heart cannot pump blood properly and supply enough to the
14
Removal of Power Line Interference and other Single Frequency Tones from Signals
body and brain. This can cause unconsciousness within second and death within
minutes. [7]
An ECG recording is important for clinical diagnosis and treatment; it is a graphical
recording of electrical impulses generated by heart. The ECG is needed to be done when
chest pain occurred such as heart attack, shortness of breath, faster heartbeats,
irregular heartbeats, high blood pressure, high cholesterol, check the hearts electrical
activity. [6]
15
Removal of Power Line Interference and other Single Frequency Tones from Signals
best solution for the removal of power line interference from ECG signal. Digital filter
has been selected to overcome this problem; there are few filtering solutions which
were examined before to manipulate the power line interference from signal which can
be divided into following categories. [10]
.
Low Pass Filters
General Notch-Rejection Filters
Adaptive Filters
Global filters
In the thesis two filtering solutions has been chosen for the removal of power line
interference, its respective harmonics and high frequency noise from Original ECG
signal. The removal of power line interference (50Hz) from ECG signal can be removed
by adaptive filtering while the harmonics and high frequency noise can be removed by
implementing general notch rejection filters.
To describe the power line interference, main cause of power line interference in
ECG Signal and to carry out a literature survey on removal of power line interference
from ECG signal.
To give solution for the removal of power line interference in ECG signal and to
investigation of different methods and to point out the best possible methodology
for the removal of power line interference.
To search for the adaptive filtering technique for removal of 50/60 Hz power line
frequency interference and to have extensive knowledge in, DSP i.e. adaptive
filtering, FIR, IIR, LMS algorithm etc
16
Removal of Power Line Interference and other Single Frequency Tones from Signals
To illustrate the suitable method or process to remove the noise, other unwanted
components, undesired tones and interferences from information signal.
To search for the best possible method available to be used for the removal of
power line interference through the internet, journals & books and analysis of
different kind of signals using MATLAB software.
In additional to investigate and analyze the harmonics and high frequency
interference in original ECG signal and to remove these noises by using general
notch rejection filters and windowed sinc low pass filter.
In the project, the deductive experimental research strategy is preferable, in this
view the investigation, observation and testing to get the desired objectives can be
accomplished on the basis of experimental works and to continue the future
research and development regarding this view.
17
Removal of Power Line Interference and other Single Frequency Tones from Signals
3 CHAPTER-3
BASIC THEORY OF DIGITAL FILTERS
3.1 Digital Signal Processing (DSP)
Digital signal processing (DSP) is well-known as compared to analog signal processing in
different applications. [13] The signals are time varying quantities which carry
information i.e. audio signal, video signals, biological signals (electrical pulses from the
heart) and communications signals. [11] A digital signal is define as the signal that has
discrete amplitude and time. These signals are represented by sequences of numbers
with finite precision and can be used when processing information by computer. [15]
The processing of the signals which deals with sound and images are known as Digital
Signal Processing. In DSP the digital signal processor can be a small microprocessor or a
large programmable digital computer, which perform the desired operations on the
input signal. [16] The application of Signal Processing has grown very fast and
implemented the advanced techniques in speech recognition, image recognition, image
enhancement, audio enhancement, noise reduction, speech & audio encoding and
storage, digital music, communication and data transmission, biometrics, biomedical
applications, radar, sonar and military applications. [13][11] Digital signal processing
systems are introduced in many different applications such as multimedia, video
recording, CD player, mobile phones, computers and modem, DSP systems acquired
famous due to their reliability, accuracy, small physical sizes and flexibility.
18
Removal of Power Line Interference and other Single Frequency Tones from Signals
The charm of the digital filters lies in the fact that digital filters are realized in the
form of piece of software so can be adapted for any required amendment with
out any increase in cost or effort on the electronic circuit.
19
Removal of Power Line Interference and other Single Frequency Tones from Signals
Plenty of support is available for design and testing of digital filters in the form of
CAD and other pre-built package on a general-purpose computer or workstation.
Realization of the digital filters can be made using the generic hardware (PC,
DAC, ADC, CPLDs) without indulging in the design of specific hardware, which
also helps in reducing the cost and time required in this process.
Digital filters can be used in biomedical instruments where frequency is very low
and analog filters are unpractical. [17]
Digital filters are most common use in any modern communication system and
digital systems are unaffected by temperature variation. [21]
With the advancement in technology the digital filters can be managed at low
frequency signals precisely. The technology of the DSP is increasing very fast, so
the digital filter are applied for the high frequency signal as well i.e. RF (radio
frequency) domain, which in the past was the limited to analog technology. [22]
Digital filters can be seen as the direct result of the advancement made in the
computers and computing technology so it will continue to benefit from the
increase in speed and power of computing chips in the future.
10 Digital filters are small in size, consume less power and can be implemented with
less cost and it also store data for future use. [17]
11 Digital filters can made accurate processing, permits the implementation of
many different operations and become more economical. [19]
20
Removal of Power Line Interference and other Single Frequency Tones from Signals
the software control, digital filter can be easily changed and are well-suited to do this
task as compared to analogue filters because of their outstanding results.
The digital filters are divided into two basic types, Finite Impulse Response (FIR) and
Infinite Impulse Response (IIR) filters. [23]- [13] - [24] which are known as non recursive
and recursive filters. [25] These types of digital filters can perform paramount required
tasks yet they are categorized either as non recursive filter known as Finite Impulse
Response (FIR) filters or a recursive filter known as Infinite Impulse Response (IIR)
filters.[26]- [25] The term non recursive filter & finite impulse response (FIR) filter and
recursive filter & infinite impulse response (IIR) filter are almost synonymous. [19]. Non
recursive that as implied by the name has a finite length impulse response and recursive
that has an impulse response of infinite length. Both types of filters can be used in the
realization of the noise cancellation but an FIR filter was chosen since it is simple and
stable. [26]
The variety of design methods has been involved for the designing of digital filters to
meet different specifications. The lowest order filter is chosen to avoid the overdesign.
The choice between the filter type (recursive and non recursive) is done due to the
computational property and the storage required for the implementation.
21
Removal of Power Line Interference and other Single Frequency Tones from Signals
First-Order Filter
The first-order filter only use the previous input and the current input is not used, so the
previous input (xn-1) is required to calculate yn. [27]
yn = xn-1
3.7.1.3 Second-Order Filter
The second-order filter compute the current output yn, two previous inputs (xn-1 and
xn-2) are needed; this is therefore called a second-order filter. [27]
yn = (xn + xn-1 + xn-2) / 3
22
Removal of Power Line Interference and other Single Frequency Tones from Signals
1/ 2min
In the equation above, is the step-size, min is the minimum eigen value, and is the
overall time-constant.
3.7.3 Coefficient of the Digital Filter.
Coefficient of the digital filter is known as tap weight, which is used to multiply against
delayed signal sample values within a digital filter structure. For an FIR filter, the filter
coefficients are the impulse response of the filter. [28]
The coefficient of the digital filter can be expressed clearly from the equations which are
described in the previous section in order of digital filter and are shown in following
table. [29]
Zero order:
yn = a0xn
First order:
yn = a0xn + a1xn-1
Second order:
In the table above the zero-order, first-order and second-order of digital filter is
presented, which shows the coefficient of the digital filter as well. The constants e.g. a0,
a1, a2 ... showing in the above equations are called the filter coefficients.
23
Removal of Power Line Interference and other Single Frequency Tones from Signals
4 CHAPTER-4
ADAPTIVE FILTERS
4.1 Introduction
The adaptive filter can be defined as, a filter which self adjust its transfer function
according to an optimizing algorithm and object can be achieve by the modification of
its characteristics. [2] Adaptive signal processing has been introduced and its growth to
the advanced related fields of digital computing, DSP and high speed integrated circuit
technology has been made rapidly. The least mean square (LMS) adaptive filtering
algorithms first paper was published in 1959 by Widrow and Hoff. [30]
Adaptive filters are extensively used in the variety of application and they had been
firstly proposed by Kelly of Bell Telephone Laboratories around 1965, [20] most of the
applications are in telecommunication for the cancellation of noise and echoes in the
transmission channel and also used in digital controller for active noise control. [31]
An adaptive filter is a digital filter whose characteristics change in an unknown
environment input signal. In the advanced era of cellular phone, digital television,
wireless communication and digital multimedia commercial services, advanced adaptive
signal processing may give the better solution for the technical problem. [30] The
adaptive filter is also used in the field of biomedical, sonar, radar and image signal
processing, telecommunication for noise cancellation etc.
24
Removal of Power Line Interference and other Single Frequency Tones from Signals
25
Removal of Power Line Interference and other Single Frequency Tones from Signals
The adaptive filters are much famous due to their economical quality, fast processing,
their short period of time adaptation and residual error is small after adaptation. [25]
Adaptive filtering is the most important technique which is used is numbers of
biomedical applications. [4]
The principal of adaptive filter is required to understand the adaptive filtering clearly
which is showed in the figure 4.1. The error signal e(n) can be generated by the output
of the programmable, variable-coefficient digital filter subtracted from a reference
signal y(n). [32]
y(n)
X(n)
PROGRAMMABLR
DIGITAL FILTER
y(n)
Input Signal
+
Reference
e(n)
ADAPTIVE ALGORITHM
FOR
COEFFICIENT UPDATING
26
Removal of Power Line Interference and other Single Frequency Tones from Signals
passed through further processing. The filters output is observed by determines its
quality for particular application. After measuring the quality it also examined by a
circuit whether it is need to improve the quality of the output signal. This processing
loop continues until the filters parameters are adjusted properly, so the filters output
quality should be as good as possible. [37]
Filter
Input
Adaptive
Filter
Filter
Adaptatio
n Rules
Filter
Outpu
t
Quality
Assessmen
t
The choice of filter structure and adaptation algorithm is important for the design
adaptive filter; the structure can be non recursive or recursive. [25]
27
Removal of Power Line Interference and other Single Frequency Tones from Signals
To satisfy the robustness of the adaptive algorithm the value of step size needs to be
small. [39] Robustness is an important criterion which is difficult to measure in a
quantitative approach. The satisfaction for the robustness of the adaptive algorithm can
be gained by the removal of external noise disturbances and arithmetic quantization
noise. [30]
[1]
The convergence ability of the LMS algorithm can be examined by the range of
convergence factors which provide the stability. [33] Faster convergence is better
solution for the allocation of additional resources in high frequency operation such as
mobile radio, cellular telephone, digital television (HDTV). In low frequency application
such as adaptive echo cancellation and audio band noise cancellation, the convergence
criterion is slow and this simple and adequate solution is provided by LMS algorithm.
[30]
The convergence performance of the LMS algorithm for FIR filter structure is controlled
by the autocorrelation matrix Rx.
Rx = E[ x * (n) x T (n)]
[30]
The condition of the satisfaction can be checked and LMS algorithms condition must be
satisfied if the step size parameter satisfies the condition. [1] The autocorrelation matrix
Rx is necessary for the convergence. The condition which is important for the
convergence criterion and the convergence factor of LMS algorithm must be chosen in
the range is
0< < 1 / max
[33]-[30]
Were max is the largest eigen value of the correlation matrix Rx. The speed of the LMS
algorithms convergence is dependent on eign value. [33] The choice of in the locality
of 1/ max is the best convergence for the adaptive algorithm. [1]-[30]
28
Removal of Power Line Interference and other Single Frequency Tones from Signals
If the matrix Rx has large eigenvalue then the vaule of must be much smaller than the
upper band. As a result the convergence speed of the coefficient will be primarily
dependent on the value of the smallest eigenvalue. [33]
noise, w(n)
+
Unknown System
H(z)
y(n)
+
input, x(n)
a(n)
Adaptive Filter
^H(z)
^y(n)
output
output d(n)
error, e(n)
Figure 4.3- System Identification Configuration for Adaptive Filter [34] -[30]-[32]
The parameters of the adaptive filters are then activated to minimize the particular
function error signal e(n). In system identification configuration when the adaptive filter
accumulated the stable values then error signal e(n) will be reduced.[30] It can be
possible for adaptive filter to converge it to a good model to match after convergence of
MSc Thesis Report, Mlardalen University, Sweden
29
Removal of Power Line Interference and other Single Frequency Tones from Signals
the unknown system by giving sufficient degree of freedom to the adaptive filter. [34][33]
The system identification configuration is the essential adaptive filtering concept. [30]
This is used to remove the error signal while processing of any input signal with noise. It
is necessary for good performance of adaptive filter to remove the noise which is mixed
with input of the system; otherwise this noise appears at the output and can decrease
the quality of the output.
30
Removal of Power Line Interference and other Single Frequency Tones from Signals
5 CHAPTER-5
ADAPTIVE ALGORITHM FOR FIR FILTERS
5.1
Introduction
The adaptive algorithm for FIR filters are is widely used in different applications such as
biomedical, communication and control due to its easily implementation, stability and
best performance. Its simplicity makes it attractive for many applications where it is
need to minimize computational requirements.
31
Removal of Power Line Interference and other Single Frequency Tones from Signals
+
+
a0
a1
-1
a2
^
y(n)
-1
an
-1
-1
32
Removal of Power Line Interference and other Single Frequency Tones from Signals
5.4.2 Disadvantages:
FIR FILTERS
Less Efficient
No Analog Equivalent
Always Stable
Linear Phase Response
Less Ringing on Glitches
CAD Design Packages Available
Decimation Increases Efficiency
33
Removal of Power Line Interference and other Single Frequency Tones from Signals
Comparison between the different characteristics of FIR and IIR can be made in variety
of ways. The most important benefit of FIR filters over IIR filters is the flexibility in
preferred frequency retort. The FIR filter might be ideal when hardware minimization
and delay minimization are important. [43] Its easy to design the non recursive filters as
compared to recursive filters. [25] Most of the applications for adaptive FIR and IIR
filtering include noise and echo cancellation and the FIR and IIR structures can be used
as adaptive filters and Stability problems. FIR filters are used for all practical
applications. [44] Reliability which is always an important issue regardless of the nature
of the design, is also an advantage offered by the FIR filters over IIR filters.
34
Removal of Power Line Interference and other Single Frequency Tones from Signals
This Adaptive FIR signal processing model can be used for different applications which
are as follow,
(a)
(b)
(c)
(d)
Noise cancellation
System identification
Inverse system identification
Prediction
35
Removal of Power Line Interference and other Single Frequency Tones from Signals
6 CHAPTER-6
LEAST MEAN SQUARE (LMS) ALGORITHM
36
Removal of Power Line Interference and other Single Frequency Tones from Signals
LMS have the ability to reject noisy data due to minute step size parameter .
LMS demonstrate slowly time varying system.
LMS algorithm does not get stuck at undesired local minima.
LMS is computationally simple memory competent.
In general LMS adaptive filter removes noise or obtains a desired signal by adapting the
filter coefficient with least-square algorithm based on given filter. [49] The performance
of the LMS algorithm is very high and it is simple in implementation for the removal of
low frequency noise. The suitable value for step size parameter can be selected
according to the applications requirement.
Primary input
Z-
Reference input
37
Removal of Power Line Interference and other Single Frequency Tones from Signals
In the figure 6.1, the overview of least mean square (LMS) algorithm is shown. The
primary input has been taken, where X is the reference input. The error signal occurs
for the desired output, there LMS adaptive filter has employed to manipulate the error.
The error signal manipulated by the adaptive algorithm is describe as
[50]
The equation above shows the desired signal and the filter output, where d(n) is the
desired signal and y(n) is the filter output. For the minimization of error signal the input
vector x(n) and e(n) are employed. Here it needs to work according to the criterion that
is supposed to minimize. The input vector is used to update the adaptive coefficients
according to that criterion. The criterion used here is the mean-square error (MSE) e:
e = E[e2(n)]
[50]
38
Removal of Power Line Interference and other Single Frequency Tones from Signals
d(n)
x(n)
Transversal
Filter
y(n|n-1)
e(n)
c(n-1)
Adaptation
Algorithm
The outline of adaptive signal processing system is describes in figure 6.2. The input to
the adaptive filter is described as scalar real-valued discrete-time signal x(n) where n is
the time index. At time n, the samples x(n), x(n - 1), . . . through x(n -N + 1) are
simultaneously present in the delay elements of the filter.
39
Removal of Power Line Interference and other Single Frequency Tones from Signals
40
Removal of Power Line Interference and other Single Frequency Tones from Signals
The Fig 6.3 describe the complete process of noise cancellation step by step
(a)
(b)
(c)
(d)
The principle of noise cancellation is to obtain noise signal and subtract it from the
corrupted signal. The required signal can be found after subtracting the corrupted signal
from noise. [36] The figure above represents the information signal (a) and the noise
signal (b), both the signals are mixed together and then mixed signal (c) is displayed. To
remove the noise signal (b) from information signal (a) there must be some filtration
method which is based on the application requirement. The original signal (d)
represents the output signal after filtration.
Signal
Source
Primary
Input
System Output
d(n)
s + no
z
Filter
Output
n1
Noise Source
e(n)
x(n)
Reference
Input
Adaptive
Filter
y(n)
Error Signal
41
Removal of Power Line Interference and other Single Frequency Tones from Signals
At the primary input signal s is measured together with noise is assumed to be the sum
of an information signal and sinusoidal interference. For monitoring of the noise n1 a
reference input supplies a correlated version of the sinusoidal interference. The
correlation of the noises no and n1 is assumed to be high and have same origin so that
influence of the useful signal s is negligible at the reference input. [31] The filter uses
the reference input to provide an estimate of sinusoidal interfering signal contained in
the primary input. The adaptive filter forms as estimate of no thus by subtracting the
adaptive filter output from the primary input signal. So the information signal with noise
is cancelled at the output by adaptive noise cancellation method. [30]
There are two important characteristics of LMS algorithm i.e. canceller behaves as an
adaptive notch filter, which is tunable and the notch in the frequency response can be
made very sharp by choosing the small value of the step size parameter . [48] The
noise cancellation is required to remove unnecessary noise from the given signal. The
term cancellation principle is used to detect the noise and subtract that noise from the
corrupted signal. Its feasibility depends upon the availability of a noise signal originating
signal. The concept of noise cancellation in its simplest form is described in figure 6.3.
Reference
Signal
Adaptive
filter
y +
-
d
e
System
Output 1
(d
Figure 6.5- Interference Cancellation by Adaptive Filtering [54]
42
Removal of Power Line Interference and other Single Frequency Tones from Signals
43
Removal of Power Line Interference and other Single Frequency Tones from Signals
7 CHAPTER-7
DESIGN, ANALYSIS AND SIMULATIONS
7.1 Main Objective
The goal was to implement and analyze of different noise removal techniques in ECG
signal. Removal of power line interference 50/60 Hz from ECG signal and how to apply
the adaptive LMS algorithm for removing of power line interference and its harmonic
from signals that has been corrupted and the original information contained by
unwanted interferences was main purpose of the thesis. The additional milestone of the
project was to investigation and implementation for the removal of harmonics and high
frequency noise from ECG signal in MATLAB environment.
Numbers of algorithms can be implemented in ECG signal to remove the different type
of noises. In this thesis, two different types for removal of noises have been
implemented.
While the thesis problem and purpose was not as simple to solve and implement as it
might, the possible benefits were enormous. The challenges for the project were the
study of most sensitive biomedical monitoring equipments such as Electrocardiogram
(ECG), electroencephalogram (EEG) and electromyography (EMG). The thesis is basically
research oriented; therefore sufficient reference materials were not available easily. The
results, analysis, implementation and conclusion of the thesis are totally depends upon
better understanding of ECG signals simulation and calculation. It was totally out of
domain so a lot of journals, research papers, internet, university library database and
books are read to analyze and understand it in a better way.
44
Removal of Power Line Interference and other Single Frequency Tones from Signals
45
Removal of Power Line Interference and other Single Frequency Tones from Signals
generated. The FIR adaptive filter has then implemented there and LMS algorithm has
been developed to filter out the low frequency noise from ECG signal. The simulation
results with graphs for ECG signal were then plotted to compare the filtered output by
putting different values of LMS step size and filter tap to visually measure the simulation
program performance. Finally the algorithm implementation and verification has been
presented which increases the reason for the selection of LMS algorithm based on FIR
filters in MATLAB environment. The affects of on convergence rate and stability for
the LMS adaptive filter has been examined to achieve the desire results.
The simulation of harmonics of high frequency noise has also been generated to find out
another filtration method in MATLAB environment. The general notch rejection filter
and windowed sinc low pass filter has been developed in MATLAB for the
implementation and removal of harmonics & high frequency noise to achieve the
required output.
7.4.1 Removal of Power Line Interference (50 Hz) from ECG Signal
by LMS Algorithm
The ECG signal has been taken and LMS adaptive filter algorithm has been developed.
The ECG signal of 50 Hz is displayed in MATLAB environment as ECG Signal and than the
noise of 50 Hz is generated and then mixed with the ECG Signal, which is displayed as
Mixed signal. The adaptive filter is implemented by using LMS algorithm, FIR filter has
been designed. The ECG Signal, 50 Hz Noise signal, Mixed signal, Error signal and
Adaptive LMS filtered output signal have been displayed. The output is nearly same as
the ECG inputted signal.
46
Removal of Power Line Interference and other Single Frequency Tones from Signals
7.4.1.1 Simulation for Power Line Interference and its Removal from ECG Signal
FilterTap=16;
mu = 0.005;
1000
2000
3000
4000
5000
6000
50 Hz Noise Signal
7000
8000
9000
10000
1000
2000
3000
4000
5000
Mixed Signal
6000
7000
8000
9000
10000
1000
2000
3000
4000
5000
Error Signal
6000
7000
8000
9000
10000
1000
2000
3000
4000
5000
6000
LMS Filtered Output
7000
8000
9000
10000
1000
2000
3000
4000
7000
8000
9000
10000
1
0
-1
2
0
-2
5
0
-5
2
0
-2
5000
6000
47
Removal of Power Line Interference and other Single Frequency Tones from Signals
FilterTap=16;
mu = 0.009;
1000
2000
3000
4000
5000
6000
50 Hz Noise Signal
7000
8000
9000
10000
1000
2000
3000
4000
5000
Mixed Signal
6000
7000
8000
9000
10000
1000
2000
3000
4000
5000
Error Signal
6000
7000
8000
9000
10000
1000
2000
3000
4000
5000
6000
LMS Filtered Output
7000
8000
9000
10000
1000
2000
3000
4000
7000
8000
9000
10000
1
0
-1
2
0
-2
5
0
-5
2
0
-2
5000
6000
48
Removal of Power Line Interference and other Single Frequency Tones from Signals
FilterTap=32;
mu = 0.005;
1000
2000
3000
4000
5000
6000
50 Hz Noise Signal
7000
8000
9000
10000
1000
2000
3000
4000
5000
Mixed Signal
6000
7000
8000
9000
10000
1000
2000
3000
4000
5000
Error Signal
6000
7000
8000
9000
10000
1000
2000
3000
4000
5000
6000
LMS Filtered Output
7000
8000
9000
10000
1000
2000
3000
4000
7000
8000
9000
10000
1
0
-1
2
0
-2
5
0
-5
2
0
-2
5000
6000
49
Removal of Power Line Interference and other Single Frequency Tones from Signals
FilterTap=32;
mu = 0.009;
1000
2000
3000
4000
5000
6000
50 Hz Noise Signal
7000
8000
9000
10000
1000
2000
3000
4000
5000
Mixed Signal
6000
7000
8000
9000
10000
1000
2000
3000
4000
5000
Error Signal
6000
7000
8000
9000
10000
1000
2000
3000
4000
5000
6000
LMS Filtered Output
7000
8000
9000
10000
1000
2000
3000
4000
7000
8000
9000
10000
1
0
-1
2
0
-2
5
0
-5
2
0
-2
5000
6000
50
Removal of Power Line Interference and other Single Frequency Tones from Signals
In the second figure ECG Signal Simulation-2, the value of filter tap and are taken as 16
and 0.009, which shows different results as compared to the ECG Signal Simulation-1. It
is concluded that the rate of convergence is changed by changing the value of .
In the figure ECG Signal Simulation-3, the value of filter tab has been changed from 16 to
32 and is taken as 0.005. This shows the ECG signal (50 Hz) and noise signals (50 Hz)
are mixed together and the filtered out by using LMS adaptive filter. This figure can be
compare with the ECG Signal Simulation-4 where the value of filter tab and are taken
as 32 and 0.009 respectively. In this simulation graph the value of is decreased which
shows the small change in the coefficient and the convergence of filter act as slowly. So
it is concluded that with large step-size the filter convergence takes place fast. It is also
concluded that with the large value of , the filter convergence act as fast. By changing
the results and convergence rates, finally it is concluded that the LMS adaptation take
place properly and it performed adaptation.
7.4.2 Removing of Harmonics and High Frequency Noise from Original ECG Signal
To the verification of removal of noise from original ECG signal, another method has
been developed for the implementation and analysis of different noise removal
technique in ECG signal. The test for the simulation of ECG signal has been taken by
adding different noises.
Humming which is also known as low frequency noise has been taken and added in the
original ECG signal. The major work is concerned with removing undesired frequency
harmonics while saving the original information in the signal as much as possible during
this elimination process.
Biomedical signals play an important role for monitoring of patients ECG. High
frequency noise can also be added with ECG signal while diagnosis of patient in hospital
or some where else. This high frequency noise can be added due to the use of Mobile
phone near patient or some other sources.
The original ECG signal with noises is passed through notch and windowed sinc low pass
filter, which filter out hum and high frequency noise portion from the ECG signal. With
this process the unwanted noise signal can be removed from ECG signal, and then
original ECG signal without noise can be obtained.
The graph of original ECG signal with hum (low frequency noise) and high frequency
noise is displayed in figure below. In the first plot of the figure the original ECG signal
has displayed and the hum and high frequency noise are displayed in second and third
plots respectively. Then both noises humming and high frequency noise has been added
to the original ECG signal, which is shown in the fourth plot as mixed signal.
51
Removal of Power Line Interference and other Single Frequency Tones from Signals
Notch rejection filter is employed for the removal of humming (50 Hz). As there are two
different noises in this case i.e. humming and high frequency noise, both noises after
passing through the notch filter, the signal with high frequency noise remain with the
original ECG signal in the fifth plot of the figure.
Finally the corrupted signal is passed through windowed sinc low pass filter to retrieve
the original ECG signal. The high frequency noise is then filtered out by using windowed
sinc low pass filter and the desire output of the ECG signal has then achieved at the final
output. By comparing the graph of input original ECG signal with the output, it is
concluded that the overall result of method and technique developed in this thesis is
achieved. The goal for the noise cancellation or removal of different noise frequencies
has been performed satisfactory.
7.4.2.1 Simulation for Harmonics and High Frequency Noise and its Removal from
Original ECG Signal
E CG
1
0
-1
200
400
600
800
1000
1200
200
400
600
800
1000
1200
200
400
600
800
1000
1200
200
400
600
800
1000
1200
200
400
600
800
1000
1200
200
400
600
800
1000
1200
HUM
1
0
Without Hum M ix ed S ig
HF-Nois e
-1
20
0
-20
20
0
-20
20
0
-20
Output
2
0
-2
Original ECG Signal with HUM and High Frequency Noise Simulation by Notch Filter
52
Removal of Power Line Interference and other Single Frequency Tones from Signals
After analysing the filtering technique for humming and high frequency noise from ECG
signal, it is concluded that implementation and analysis has been performed more or
less the same and it also shows the method and filtering technique developed for the
removal of harmonics and high frequency noise from ECG signal performed its operation
and meet the required results.
53
Removal of Power Line Interference and other Single Frequency Tones from Signals
frequency noise of 375 Hz. After passing the ECG signal with both noises through notch
filter, the ECG signal with high frequency noise remains. Then this remaining signal with
high frequency noise has been passed through the windowed sinc low pass filter to get
rid of high frequency noise. By use this method the original ECG signal can be retrieved.
By investigating this method, it is concluded that the overall result of the technique is
achieved. The main goal for the removal of different noise i.e. harmonics and a high
frequency noise has been performed satisfactory.
54
Removal of Power Line Interference and other Single Frequency Tones from Signals
8 CHAPTER-8
8.1 CONCLUSION AND FUTURE RESEARCH
8.1.1 Conclusion
This report is devoted to the problems and solutions on removal of Power Line
Interference and other Single Frequency Tones from Signals. The understanding of noise
cancellation from ECG signal was explained clearly to the readers, the methods and
techniques applicable to be used discussed through out the report. It has been
proposed a solution for the power line interference its respective harmonics and high
frequency noise interferences from original ECG signal. The results have been obtained
which were required in purpose statement of the report.
In general FIR filters are used because these types of filters have simple architecture and
are logically stable, so the FIR filter was chosen for the development of the system. The
research has been made for the selection of filters and algorithm, two kind of adaptive
filters i.e. FIR (Finite impulse response) & IIR (Infinite impulse response) and two kinds
of algorithms, the LMS (least mean square) & RLS (recursive least squares) algorithms
were studied and examined. The LMS algorithm is the most widely used adaptive
filtering algorithm in biomedical monitoring equipments, so it was decided to be
employed for the thesis to get the required solution for the main purpose. The
applications of the LMS algorithm can be implemented due to its simplicity and
robustness.
An adaptive filter is used in applications that require different filter characteristics in
response to variable signal conditions. The speed of adaptation and accuracy of the
noise cancellation after adaptation are important measures of performance for noise
cancellation algorithm. The goal of the adaptive filter is to match the filter coefficients
to the noise so that the adaptive filter can subtract the noise out from signal. So
adaptive filtering technique was selected to achieve the goal of thesis.
The test for the simulation of ECG signal has been taken. The signal is corrupted by
power line interference of 50 Hz. It is observed that the frequency of the power line
interference is 50 Hz which is then mixed with original ECG signal, it is also examined
that the mixed signal is displayed on the plot. After passing through LMS algorithm the
filtered output is nearly same as the input signal with some acceptable distortion range.
The value of step size play an important role in determining the convergence speed,
stability and residual error after convergence. The convergence rate was controlled by
55
Removal of Power Line Interference and other Single Frequency Tones from Signals
LMS step size . The ECG signal graphs described in the simulation results verify the
adaptation of the LMS adaptive algorithm by changing various parameters like step size,
convergence value () and filter taps have various effects on the output graphs. The
result shows that LMS is an effective algorithm used for the adaptive filter in the noise
cancellation implementation.
By increasing the filter order it shows a convergence rate but makes the results more
precise and by decreasing the step size value it creates the slower convergence but
improves the stability and accuracy. The recovered signal closely resembles to the
original simulated signal minus the noise. It can be seen that the implementation of the
algorithm functions as correctly and efficiently. By comparing the graphs of the input
signal of ECG and output signal, it is noticed that the simulation program performs
satisfactorily and that noise cancellation from original ECG signal is acquired. The overall
performance of LMS algorithm for power line interference is achieved.
Furthermore the general notch rejection filters method also performs the correct
operation while filtering the noise from original ECG signal. This technique for the
investigation, implemented and analysis of removal of harmonics and high frequency
noise from original ECG signal performed satisfactory. It is concluded that the low
frequency noise (hum) and high frequency noise can be removed from original ECG
signal by the implementation of general notch rejection filters method and the desired
result can be achieved accurately.
8.1.2 Suggestion for Future Work/Future Enhancement
The depth knowledge achieved is in a number of aspects by using digital signal
processing techniques with MATLAB package for medical monitoring equipments (ECG).
It provides the real concepts along with the theoretical backgrounds of removal of
power line interference, single frequency tones and high frequency noise from original
ECG signal. This enhances the understanding and self confidence in the field of
electronics and biomedical engineering.
In the thesis, the adaptive signal processing filtering technique based on LMS algorithm
could be implemented for more signals and also improvement of the thesis can be
further implemented with different algorithms such as NLMS and RLS to achieve the
desired results. It could also be investigate and implement the removal of multiple of
harmonics from ECG signal.
In further research it would be of interest to make a broader study and look at some
companies for related project for the implementation.
56
Removal of Power Line Interference and other Single Frequency Tones from Signals
9 APPENDIX A
clear;
clc
close all;
fullscreen = get(0,'ScreenSize');
figure('Position',[ fullscreen(3)/4 fullscreen(4)/4 fullscreen(3)/1.5 fullscreen(4)/1.5])
ECG_signal=ecg(50);
ECG=ECG_signal;
for i=1:Length/50-1
ECG_signal=[ ECG_signal ECG];
clc
end
57
Removal of Power Line Interference and other Single Frequency Tones from Signals
frequency1=50;
Noise_signal = sin(2*pi*frequency1*i/Fs);
x=ECG_signal;
n=Noise_signal;
m=Mixed_signal;
d=x+n+m;
n1=rand()*Noise_signal;
length_filter=15;
w=ones(1,length_filter);
mu=0.005;
FilterTap=16;
for i=1:Length
y1=0;
if i<length_filter
for j=1:i
y1=y1 + w(i,j)*n1(i-j+1);
end
else
58
Removal of Power Line Interference and other Single Frequency Tones from Signals
for j=1:length_filter
y1=y1+ w(i,j)*n1(i-j+1);
end
end
y(i)=y1;
e(i)=d(i)-y(i);
if i<length_filter
for k=1:i
w(i+1,k)=w(i,k) + mu*n1(i-k+1)*e(i); % mu*n1(i)*e(i) ; %
end
else
for k=1:length_filter
w(i+1,k)=w(i,k) + mu*n1(i-k+1)*e(i);
end
end
end
Output_without_Noise_signal=e;
M=100;
fc=0.25;
for i=1:M
h(i)=sin(2*pi*fc*(i-M/2))*(0.42- 0.5*cos(2*pi*i/M) + 0.08*cos(4*pi/M) )/(i-M/2);
end
h(ceil(M/2))=2*pi*fc;
h=h/sum(h);
Final_filter_Output=filter(h,1,Output_without_Noise_signal);
% Display ECG Signal, Noise, ECG Plus Noise and Adaptive LMS Output Signals
% Input is the ECG Signal of frequency 50 Hz
figure(1)
59
Removal of Power Line Interference and other Single Frequency Tones from Signals
60
Removal of Power Line Interference and other Single Frequency Tones from Signals
10 APPENDIX C
10.1 MATLAB CODE FOR HARMONICS AND HIGH FREQUENCY NOISES
FROM ECG SIGNAL BY GENERAL NOTCH REJECTION FILTER
% Removal of HUM and High Frequency Noise from Original ECG Signal by Notch Filter
clear;
clc
close all;
fullscreen = get(0,'ScreenSize');
figure('Position',[ fullscreen(3)/4 fullscreen(4)/4 fullscreen(3)/1.5 fullscreen(4)/1.5])
% Create the signal
% Original ECG signal
Fs = 800;
ECG_signal=ecg(50);
ECG=ECG_signal;
for i=1:20
ECG_signal=[ ECG_signal ECG];
clc
end
scrsz=get(0,'ScreenSize');
maxWindow=[1 31 scrsz(3) scrsz(4)-104];
frequency1=50;
frequency2=375;
61
Removal of Power Line Interference and other Single Frequency Tones from Signals
i=1:length(ECG_signal);
HUM = sin(2*pi*frequency1*i/Fs);
Fn = Fs/2;
W0 = 50;
w0 = W0*pi/Fn;
BandWidth = 10;
B = BandWidth*pi/Fn;
M=100;
fc=0.25;
for i=1:M
h(i)=sin(2*pi*fc*(i-M/2))*(0.42- 0.5*cos(2*pi*i/M) + 0.08*cos(4*pi/M) )/(i-M/2);
end
h(ceil(M/2))=2*pi*fc;
h=h/sum(h);
62
Removal of Power Line Interference and other Single Frequency Tones from Signals
Final_filter_Output=filter(h,1,Output_without_Hum);
figure(1)
% Filtered Output
63
Removal of Power Line Interference and other Single Frequency Tones from Signals
subplot(6,1,6);
plot(Final_filter_Output)
ylabel ('Output')
%grid on;
hold on;
64
Removal of Power Line Interference and other Single Frequency Tones from Signals
11 REFERENCES
The list of information gathered from books, journals and internet sites.
[1] Stacy Finlay, Carrie Klekta and Ernie Packulak, Adaptive Noise Cancellation for
ECG Signal, 2002
[2] http://en.wikipedia.org/wiki/Adaptive_filter
Accessed on 2008-11-22
[3] Rolf Limacher, Removal of power line interference from the ECG signal by an
Adaptive digital filter, ETH Zurich, laboratory of Electrical Engineering Design
Gloriast, Zurich
[4] Applications of Adaptive Filtering to ECG Analysis, Noise Cancellation and Arrhythmia
Detection Nitish V . Thakor, Seiiior Member, IEEE, and Yi-Sheng Zhu, Se~iior Momher,
IEEE
[5] http://www.eng.nus.edu.sg/EResnews/0410/sf/sf_7.html Accessed on 2008-11-23
[6] http://www2.netdoctor.co.uk/health_advice/examinations/ecg.htm Accessed on
2008-09-13
[7] http://www.essortment.com/all/howhumanheart_roua.htm Accessed on 2008-11-23
[8] http://www4.ncsu.edu/~xzhang11/PDF/525_proj1.pdf
[9] Dr. A. Michael Noll, Technological Challenges of Power Line, Feb 5 (2002)
California, Telecommunication, Annenberg School of Communication at the
University of Southern Los Angeles. and Baseline Drift Removal, Depart of Psychology,
Second University of Naples,Italy and International Institute for Advanced Scientific
Studies, Italy. Block Approach Harvey Mudd College, California.
[10] Anna Esposito, Pierluigi D. Andria, An Adaptive Learning Algorithm for ECG Noise
[11] John Leis, Digital Signal Processing- A MATLAB-Based Tutorial Approach, Research
Studies Press Ltd. 2002
[12] http://www.mathworks.com Accessed on
2008-11-21
[13] Hakan Johansson, Synthesis and Realization of High-Speed Recursive Digital Filters,
University of Linkoping, Sweden, 1998
[14] MATLAB Software help
[15] Philip D. Cha , John I. Molinder , Fundamentals of Signals and Systems, A Building
[16] John G. Proakis, Dimitris G. Manolakis Digital Signal Processing, Principles,
Algorithm and Applications
[17] Emmanuel. C. Ifeachor, Barrie. W. Jervis, Digital Signal Processing
[18] Rahman Jamal, Mike Cerna, John Hanks Designing Filters Using the Digital Filter
Design Toolkit
[19] Edward P. Cunningham, Digital Filtering An Introduction, John Wiley & Sons, Inc.
1995
[20] Haykin S, Adaptive Filter Theory, 2nd Edition, United Stated of American: PrenticeHall, Inc. 1991,
[21] Krister Landernas, Implementation of Digital Filters Using Digital-Series
Arithmetics, Department of electronics, Licentiate thesis No. 23, 2004
65
Removal of Power Line Interference and other Single Frequency Tones from Signals
[22] http://www.dsptutor.freeuk.com/dfilt2.htm
Accessed on 2008-11-22
[23] http://www.intersil.com/data/AN/an9603.pdf
Accessed on 2008-11-21
[24] Nicholas John Loy, An Engineers Guide to FIR Digital Filters, Prentice Hall, 1988
[25] Andreas Antoniou, Digital Filters Analysis, Design and Applications, Second Edition,
McGraw-Hill, Inc. 1993
[26] B. Widow, "Adaptive noise cancelling: principles and applications", Proceedings of
the IEEE, vol. 63,
[27] http://www.dsptutor.freeuk.com/dfilt5.htm
Accessed on 2008-11-23
[28] http://www.dspguru.com/info/terms/filtterm/index2.htm Accessed on 2008-11-22
[29] http://www.dsptutor.freeuk.com/dfilt6.htm
Accessed on
2008-11-23
[30] W. Kenneth Jenkins, Andrew W. Hull, Jeffrey C. Strait, Bernard A. Schnaufer, Xiaohui
Li, Advanced Concept in Adaptive Signal Processing, Kluwer Academic Publisher 1996
[31] Per Sjosten, Adaptive Noise Control Using The LMS Algorithm, Chalmer University
of Technology
[32] Maurice G. Bellanger, Adaptive Digital Filters, Second Edition Revised and
expended, Marcel Dekker, Inc. 2001
[33] Paulo S.R. Diniz, Adaptive Filtering Algorithm and Practical Implementation,
Kluwer Academic Publishers, 1997
[34] Michael S. White and Stuart J. Flockton, Adaptive Recursive Filtering using
Evolutionary Algorithms. Department of Physics, Royal Holloway, University of
London.
[35] B. Widrow and S. Stearns. Samuel D. Adaptive Signal Processing. Prentice Hall,
Inc,1985
[36] Farhang Boroujeny, B. Adaptive Filters: Theory and Applications. Chichester: John
Wiley and Sons, 1998
[37] Treichler, John. R. and Johnson Jr., Richard C. and Larimore, Michael G. Theory
and Design of Adaptive Filters. New York: John Wiley and Sons, 1987
[38] Yuu-Seng Lau, Zahir M. Hussian and Richard Harris, Performance of Adaptive
Filtering Algorithms: A Comparative Study, Centre for Advanced Technology in
Telecommunications (CATT), School of Electrical and Computer Engineering RMIT
University, Melbourne, Victoria, Australia.
[39] Ali H. Sayed, Fundamentals of Adaptive Filtering, John Wiley & Sons Inc, 2003
[40] Cedric Keip, Design of a Finite Impulse Response Filter Coefficients Computation
Tool Using Evolutionary Strategies and Extension to Filter Bank Design, 2004
[41] Terrell, Trevor J. Introduction to Digital Filters. Second Edition, London:
Macmillan, 1980
[42] C.F.N. Cowan, P.M Grant, Adaptive Filters, United States of America: Prentice
Hall, Inc. 1985
[43] Thomas J. Cavicchi, John, Digital Signal Processing, Wiley & Sons, Inc 2000
[44] http://dualist.stanford.edu/~ee265/www-Sept2006/docs/digital_filters.pdf
Accessed on 2008-11-22
[45] N. Kalouptsidis. Adaptive System Identification and Signal Processing Algorithm
(University of Athens) and S. Theodoridis (University of Patras) Prentice Hall
Inc.,1993.
66
Removal of Power Line Interference and other Single Frequency Tones from Signals
[46] L. Guo, L. Ljung, G.J. Wang, Necessary and Sufficient Condition for Stability of LMS,
Department of Electrical Engineering, Linkoping University, Sweden, 1995
[47] Monson H. Hayes, Jonson. Statistical Digital Signal Processing and Modeling,
Wiley & Sons.
[48] Simon Haykin, Adaptive Filter Theory, Fourth Edition, Prentice Hall, Inc 2002
[49] International Journal of Control, Automation, and Systems, vol. 3, no. 1, March
2005
[50] Charles Le and Scott Hensley. RFI removal from AIRSAR polarimetric data, Jet
Propulsion Laboratory California Institute of Technology
[51] Special issue on Adaptive Signal Processing of the International Journal computers
& Electrical Engineering Vol. 18, No. 3/4, 1992.
[52] A. Zaknich, Principles of Adaptive Filters and Self-Learning Systems, Springer
[53] John L. Semmlow, Robert Brunswick, Biosignal and Biomedical Image Processing
Matlab-Based Applications, New Hersey, U.S.A. Rutgers University, Piscataway,
New Jersey, U.S.A.
[54] Haykin, S. Adaptive Filter Theory. Third Edition. New Jersey: Prentice-Hall,
Inc.1996
67