You are on page 1of 69

DUAL TREE COMPLEX WAVELET TRANSFORM BASED EEG

DENOISING SYTEM
A Project Report Submitted To
JAWAHARLAL NEHRU TECHNOLOGICAL UNIVERSITY, KAKINADA
In partial fulfilment of the requirements for the award of degree of
Bachelor Of Technology
in
ELECTRONICS AND COMMUNICATIONENGINEERING
by

M.MANASA (15K61A0467)
N. JAYA SURYA PRAKASH KUMAR (16K65A0413)
M. ANJANAMALLIKA (15K61A0470)
M.SRI HARSHITHA (15K61A0472)
Under the esteemed guidance of
Mr CH. BABU,
Assistant Professor

Department of Electronics & Communication Engineering


SASI INSTITUTE OF TECHNOLOGY&ENGINEERING
Kadakatla, TADEPALLIGUDEM– 534 101
Academic Year 2018-19
SASI INSTITUTE OF TECHNOLOGY&ENGINEERING
Department of Electronics and Communication Engineering
Kadakatla, TADEPALLIGUDEM– 534 101
Academic Year 2018-19

CERTIFICATE

This is to certify that the project work entitled Dual tree Complex Wavelet
Transform Based EEG Denoising System is being submitted by M Manasa
(15K61A0467), NJ Surya Prakash Kumar (16K65A0413), M AnjanaMallika
(15K61A0470), M Sri Harshitha (15K61A0472)in partial fulfillment for the award
of Degree of BACHELOR OF TECHNOLOGY in ELECTRONICS &
COMMUNICATION ENGINEERING to the Jawaharlal Nehru Technological
University, Kakinada during the academic year 2018-19 is a record of bonafide
work carried out by them under our guidance and supervision.

Project Guide Head of Department


Mr.Ch.Babu, Mr.T.J.V.Subhramanyam,
Assistant Professor Professor & HoD

External Examiner
DECLARATION

We, M Manasa (15K61A0467), N J Surya Prakash Kumar (16K65A0413), M Anjana

Mallika (15K61A0470), M Sri Harshitha (15K61A0472), hereby declare that this thesis titled

“Dual tree Complex Wavelet Transform Based EEG Denoising System” under the guidance

and supervision of Mr. Ch. Babu, Assistant Professor, ECE Department, Sasi Institute of

Technology & Engineering, Tadepalligudem., is a bonafied research work submitted in partial

fulfilment of the requirements for the award of the degree of Bachelor of Technology. The work

carried out by them and results embodied in this thesis have not been reproduced or copied from

any source.

We also declare that it has not been submitted previously in part or in full to this

university or any other university / Institution for the award of any degree or diploma.

Place: Tadepalligudem
Date: _______________

With gratitude,

1. M Manasa (15K61A0467)
2. N J Surya Prakash Kumar (16K65A0413)
3. M Anjana Mallika (15K61A0470)
4. M Sri Harshitha (15K61A0472)
ACKNOWLEDGMENT

We take immense pleasure to express our deep sense of gratitude to our beloved Guide
Mr. Ch Babu, Assistant Professor, ECE Department, Sasi Institute of Technology&
Engineering, Tadepalligudem-534101, for her valuable suggestions and rare insights, constant
encouragement and inspiration throughout the project work.

We express our deep sense of gratitude to our beloved Principal, Dr. K.Bhanu Prasad,
Sasi Institute of Technology& Engineering, Tadepalligudem-534101, for his valuable guidance
and for permitting us to carry out this project.

We would like to take this opportunity to thank Dr. N Venkata Rao, Dean Academics,
Sasi Institute of Technology & Engineering, Tadepalligudem-534101, for providing a great
support in successful completion of our project.

We express our deep sense of gratitude to Dr. T. J. V. Subrahmanyeswara Rao, HOD,


ECE Department, Sasi Institute of Technology& Engineering, Tadepalligudem-534101, for
the valuable guidance and suggestions, keen interest shown thorough encouragement extended
throughout the period of project work.

We are grateful to my project coordinator and thanks to all teaching and non teaching
staff members those who contributed for the successful completion of our project work.

With gratitude,

1. M Manasa (15K61A0467)
2. N J Surya Prakash Kumar (16K65A0413)
3. M Anjana Mallika (15K61A0470)
4. M Sri Harshitha (15K61A0472)
CONTENTS

VISION AND MISSION i

PEO’ AND PO’S ii

ABSTRACT iii

LIST OF FIGURES iv

LIST OF TABLES v

NOMENCLATURE vi

CHAPTER1: INTRODUCTION
1.1. The electroencephalogram 2
1.1.1 Source of EEG activity 2
1.1.2 Clinical use 3
1.1.3 Research use 5
1.1.4 EEG Recording Method 6
1.1.5 Normal activity 9
1.1.6 Comparison table 10
1.1.7 Wave patterns 11
1.2 Artifacts 14
1.2.1 Biological artifacts 14
1.2.2 Environmental artifacts 16
CHAPTER2: LITERATURE SURVEY 19
CHAPTER3: WAVELET AND EEG 26
3.1 Wavelet transform 26
3.1.1 Time domain Features 27
3.1.2 Frequency domain features 28
3.1.3 Wavelet domain features 28
3.1.4 Wavelet Family 29
3.2 Wavelet analysis 29
3.3 Wavelet Decomposition 29
3.4 Wavelet Multiresolution Analysis 30
CHAPTER4: METHODOLOGY 33
4.1 Introduction 33
4.2 Principle Component Analysis 36
4.3Blind Source Separation 36
4.4 Linear Filtering 37
4.5 Independent Component Analysis 38
4.6 Proposed Method 39
4.6.1 EEG Recording 39
4.6.2 Procedure for wavelet multiresolution analysis 40
4.6.3 ICA decomposition 40
4.6.4 Wavelet artefact removal 41
4.6.5 Wavelet and ICA reconstruction 42
CHAPTER 5: RESULTS 43

5.1 Experiments and Results 43

6. CONCLUSION AND FUTURE WORK 46

7. REFERENCES 47

APPENDIX – A 48

APPENDIX – B 49

APPENDIX – C 53
A Project Report on Dual Tree Complex Wavelet Transform based EEG Denoising System

VISION & MISSION


INSTITUTE VISION:

Confect as a premier institute for professional education by creating technocrats who


can address the society’s needs through inventions and innovations.

INSTITUTE MISSION:

• Partake in the national growth of technological, industrial area with societal


responsibilities.
• Provide an environment that promotes productive research.
• Meet stakeholder’s expectations through continued and sustained quality
improvements.

DEPARTMENT VISION:

To help in making the institute in providing competitive engineering education to the


learner and bring out quality professionals in the field of Electronics and Communication
Engineering, who can meet the industrial needs by taking up existing, new engineering and
social challenges.

DEPARTMENT MISSION:

• To provide quality and effective training in the domain of Electronics and


Communication Engineering through curriculum, effective teaching, and
learning process.
• Provide state of art laboratories.
• Conduct industrial collaborative programs.
• Involve the stakeholders in Co-curricular & extracurricular activities.

Sasi Institute of Technology and Engineering Page i


A Project Report on Dual Tree Complex Wavelet Transform based EEG Denoising System

PROGRAM EDUCATIONAL OBJECTIVES


PEO1. Develop strong foundation in Electronics and Communication Engineering to achieve
the needs of industry with continuous skill improvement.

PEO2.Contribute to society in solving technical problems using electronic and


communication principles, tools, practices and Team work.

PEO3. Personally encourage to uphold to professional, ethical, social, environmental


responsibilities of their profession.

PROGRAM OUTCOMES & PROGRAM SPECIFIC


OUTCOMES
PROGRAM OUTCOMES:
PO1: Engineering knowledge

Apply the knowledge of mathematics, science, engineering fundamentals and an


engineering specialization to the solution of complex engineering problems.

PO2: Problem analysis

Identify, formulate, research literature and analyze complex engineering problems


reaching substantiated conclusions using first principles of mathematics, natural sciences and
engineering sciences.

PO3: Design/development of solutions

Design solutions for complex engineering problems and design system components or
processes that meet specified needs with appropriate consideration for public health and
safety, cultural, societal and environmental considerations.

PO4: Conduct investigations of complex problems

Use research based knowledge and research methods including design of experiments,
analysis and interpretation of data and synthesis of information to provide valid conclusions.

PO5: Modern tool usage

Create, select and apply appropriate techniques, resources and modern engineering
and IT tools including prediction and modelling to complex engineering activities with an
understanding of the limitations.

Sasi Institute of Technology and Engineering Page ii


A Project Report on Dual Tree Complex Wavelet Transform based EEG Denoising System

PO6: The engineer and society

Apply reasoning informed by contextual knowledge to assess societal, health, safety,


legal and cultural issues and the consequent responsibilities relevant to professional
engineering practice.

PO7: Environment and sustainability

Understand the impact of professional engineering solutions in societal and


environmental contexts and demonstrate knowledge of and need for sustainable development.

PO8: Ethics

Apply ethical principles and commit to professional ethics and responsibilities and
norms of engineering practice.

PO9: Individual and team work

Function effectively as an individual, and as a member or leader in diverse teams and


in multidisciplinary settings

PO10: Communication

Communicate effectively on complex engineering activities with the engineering


community and with society at large, such as being able to comprehend and write effective
reports and design documentation, make effective presentations and give and receive clear
instructions

PO11: Project management and finance

Demonstrate knowledge and understanding of engineering and management


principles and apply these to one’s own work, as a member and leader in a team, to manage
projects and in multidisciplinary environments

PO12: Life-long learning

Recognize the need for and have the preparation and ability to engage in independent
and life- long learning in the broadest context of technological change.

PROGRAM SPECIFIC OUTCOMES

PSO1: Practice in Embedded Systems

An ability to recognize and adapt to emerging trends in embedded systems and its
applications

Sasi Institute of Technology and Engineering Page ii


A Project Report on Dual Tree Complex Wavelet Transform based EEG Denoising System

PSO2: Signal & Image Analysis

An ability to perform Signal & Image processing in the field of communication

PSO3: Digital System design

An ability to design a system or process to meet desired needs in VLSI.

Sasi Institute of Technology and Engineering Page ii


A Project Report on Dual Tree Complex Wavelet Transform based EEG Denoising System

ABSTRACT
Brain electrical activity recordings by electroencephalography (EEG) are often
contaminated with signal artifacts. Procedures for automated removal of EEG artifacts are
frequently sought for clinical diagnostics and brain computer interface (BCI) applications. In
recent years, a combination of independent component analysis (ICA) and discrete wavelet
transform (DWT) has been introduced as standard technique for EEG artifact removal.
However, in performing the wavelet-ICA procedure, visual inspection orarbitrary
thresholding may be required for identifying artifactual components in the EEG signal.

In this project we proposed a novel approach for identifying artifactual components


separated by wavelet-ICA. This project presents a robust and extendable system that enables
fully automated identification and removal of artifacts from EEG signals, without applying
any arbitrary thresholding. Using test data contaminated by eye blink artifacts, our method
performed better in identifying artifactual components than existing thresholding methods.
Wavelet-ICA successfully removed target artifacts, while largely retaining the EEG source
signals of interest. This combinatorial method is also extendable to accommodate multiple
types of artifacts present in multi-channel EEG.

EXPECTED OUTCOMES:

PO1: Engineering Knowledge

PO2:Problem analysis

PO3:Design/Development of solutions

PO4:Conduct investigations of complex problems

PO5: Modern Tool usage

PO6: The Engineer and society

PO8: Ethics

PO9: Individual and Team work

PO10: Communication

PSO2: Signal and Image Analysis


Sasi Institute of Technology and Engineering Page iii
A Project Report on Dual Tree Complex Wavelet Transform based EEG Denoising System

LIST OF FIGURES

Figure

1.1 Delta waves 11


1.2 Theta waves 11
1.3 Alpha waves 12
1.4 Sensor motor rhythm 12
1.5 Beta waves 13

1.6 Gamma waves 13

1.7 EEG Signal Processing 17

3.1 Three- level wavelet decomposition tree 30

4.1 Block diagram of the proposed artifacts removal system 37

5.1 Raw EEG signal 42

5.2 Raw EEG signal classification 43

5.3 Filtered ECG signal 44

5.4 Noise added EEG signal 44

5.5 Classification of noise added EEG signal 45

5.6 Noise eliminated EEG signal 45

Sasi Institute of Technology and Engineering Page iv


A Project Report on Dual Tree Complex Wavelet Transform based EEG Denoising System

LIST OF TABLES

Table Number

1.1 Comparison of EEG bands 9

1.2 Brain Sensing Technologies 15

Sasi Institute of Technology and Engineering Page v


A Project Report on Dual Tree Complex Wavelet Transform based EEG Denoising System

NOMENCLATURE

ACRONYM ABREVIATION
EEG Electroencephalogram

DTCWT Dual Tree Complex Wavelet Transform

PCA Principle Component Analysis

BSS Blind Source Separation

ICA Independent Component Analysis

LF Linear Filtering

MRI Magnetic Resonance Imaging

ECG Electrocardiogram

FMRI Functional Magnetic Resonance Imaging

ERP Event Related Potential

DFT Discrete Fourier Transform

BCI Blunt cardiac injury.

EMD Eergency medical dispatcher

TFD Tansperent Film Dressing

Sasi Institute of Technology and Engineering Page vi


A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

Chapter 1
INTRODUCTION
Electroencephalography (EEG) is a medical technology that is used in the monitoring
of the brain and diagnosis of many neurological illnesses. EEG is the technology of choice in
epilepsy and neonatal seizure detection as well as in other diagnostics such as sleep analysis.
Similarly, in evoked event-related potentials the EEG is used to evaluate brain function, often
in patients with cognitive diseases. In addition, many brain-computer interface (BCI)
applications utilize EEG as a direct communication pathway between the brain and an
external device, most commonly for assisting, augmenting, or repairing human cognitive or
sensory-motor functions.
To utilize the EEG for any of the applications requires interpretation and processing
of vast quantities of information. Traditionally, EEG data is examined by a trained clinician
who identifies neurological events of interest. However, recent advances in signal are
processing and machine learning techniques have allowed the automated detection of
neurological events for many medical applications. By automating the detection of
neurologically relevant events, the burden of work on the clinician can be significantly
reduced, improving the response time to the illness, and allowing suitable medical treatment
to be administered within minutes rather than hours. In the case of BCI, automated
neurological event detection has made possible this emerging engineering field, with new
technologies and applications being created on an ongoing basis.
However, as typical EEG signals are of the order of micro volts (µV), contamination
by non-cerebral signals is frequent. These artifacts can significantly distort the EEG signal,
making its interpretation difficult, and can dramatically disapprove automatic neurological
event detection classification performance. In particular, contamination of EEG signals by
artifacts arising from head movements have been a serious obstacle in the deployment of
automatic neurological event detection systems in ambulatory EEG, i.e. environments where
the patient or user has unrestricted movement.
Similarly, analysis of epileptic and neonatal seizure detection systems developed by the
Biomedical Signal Processing Group at University College Cork (UCC), have identified
movement, ocular and respiratory artifacts as problematic, leading to a large number of false
detections, and electively preventing these automatic neurological event detection systems

Sasi Institute of Technology and Engineering Page 1


A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

from being deployed in a clinical setting. This thesis, therefore, investigates and develops
number of promising artefact detection and removal algorithms.
1.1 THE ELECTROENCEPHALOGRAM
Electroencephalography (EEG) is the recording of electrical activity along the
scalp produced by the firing of neurons within the brain. In clinical contexts, EEG refers to
the recording of the brain's spontaneous electrical activity over a short period of time, usually
20–40 minutes, as recorded from multiple electrodes placed on the scalp. In neurology, the
main diagnostic application of EEG is in the case of epilepsy, as epileptic activity can create
clear abnormalities on a standard EEG study. A secondary clinical use of EEG is in the
diagnosis of coma, encephalopathies, and brain death. EEG used to be a first-line method for
the diagnosis of tumours, stroke and other focal brain disorders, but this use has decreased
with the advent of anatomical imaging techniques such as MRI and CT. Derivatives of the
EEG technique include evoked potentials (EP), which involves averaging the EEG activity
time-locked to the presentation of a stimulus of some sort (visual, soma to sensory, or
auditory). Event-related potentials refer to averaged EEG responses that are time-locked to
more complex processing of stimuli; this technique is used in cognitive science, cognitive
psychology, and psycho physiological research.
1.1.1 Source Of EEG Activity
The electrical activity of the brain can be described in spatial scales from the currents
within a single dendrite spine to the relatively gross potentials that the EEG records from the
scalp, much the same way that economics can be studied from the level of a single
individual's personal finances to the macro-economics of nations. Neurons, or nerve cells, are
electrically active cells that are primarily responsible for carrying out the brain's functions.
Neurons create action potentials, which are discrete electrical signals that travel down axons
and cause the release of chemical neurotransmitters at the synapse, which is an area of near
contact between two neurons.
This neurotransmitter then activates a receptor in the dendrite or body of the neuron
that is on the other side of the synapse, the post-synaptic neuron. The neurotransmitter, when
combined with the receptor, typically causes an electric current within the dendrite or body of
the post-synaptic neuron. Thousands of post-synaptic currents from a single neuron's
dendrites and body then sum up to cause the neuron to generate an action potential.

Sasi Institute of Technology and Engineering Page 2


A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

This neuron then synapses on other neurons, and so on. EEG reflects correlated
synaptic activity caused by post-synaptic potentials of cortical neurons. The ionic currents
involved in the generation of fast action potentials may not contribute greatly to the averaged
field potentials representing the EEG .More specifically, the scalp electrical potentials that
produce EEG are generally thought to be caused by the extracellular ionic currents caused by
dendrite electrical activity, whereas the fields producing magneto encephalographic signals
are associated with intracellular ionic currents.
The electric potentials generated by single neurons are far too small to be picked by
EEG or MEG. EEG activity therefore always reflects the summation of the synchronous
activity of thousands or millions of neurons that have similar spatial orientation. Because
voltage fields fall off with the square of the distance, activity from deep sources is more
difficult to detect than currents near the skull. Scalp EEG activity shows oscillations at a
variety of frequencies. Several of these oscillations have characteristic frequency ranges,
spatial distributions and are associated with different states of brain functioning (e.g., waking
and the various sleep stages). These oscillations represent synchronized activity over a
network of neurons. The neuronal networks underlying some of these oscillations are
understood (e.g., the thalamocortical resonance underlying sleep spindles), while many others
are not (e.g., the system that generates the posterior basic rhythm). Research that measures
both EEG and neuron spiking finds the relationship between the two is complex with the
power of surface EEG only in two bands that of gamma and delta relating to neuron spike
activity.
1.1.2 Clinical Use
A routine clinical EEG recording typically lasts 20–30 minutes (plus preparation
time) and usually involves recording from scalp electrodes.
Routine EEG is typically used in the following clinical circumstances:
• To distinguish epileptic seizures from other types of spells, such as psychogenic non-
epileptic seizures, syncope (fainting), sub-cortical movement disorders and migraine
variants.
• To differentiate "organic" encephalopathy or delirium from primary psychiatric
syndromes such as catatonia.
• To serve as an adjunct test of brain death.

Sasi Institute of Technology and Engineering Page 3


A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

• To prognosticate, in certain instances, in patients with coma.


• To determine whether to wean anti-epileptic medications.
At times, a routine EEG is not sufficient, particularly when it is necessary to record a
patient while he/she is having a seizure. In this case, the patient may be admitted to the
hospital for days or even weeks, while EEG is constantly being recorded (along with time-
synchronized video and audio recording).
A recording of an actual seizure (i.e., an octal recording, rather than an inter-octal
recording of a possibly epileptic patient at some period between seizures) can give
significantly better information about whether or not a spell is an epileptic seizure and the
focus in the brain from which the seizure activity emanates. Epilepsy monitoring is typically
done:
• To distinguish epileptic seizures from other types of spells, such as psychogenic non-
epileptic seizures, syncope (fainting), sub-cortical movement disorders and migraine
variants.
• To characterize seizures for the purposes of treatment.
• To localize the region of brain from which a seizure originates for work-up of possible
seizure surgery.

• To monitor the depth of anaesthesia.


• As an indirect indicator of cerebral perfusion in carotid endarterectomise.
• To monitor an barbital effect during the Wada test.
EEG can also be used in intensive care units for brain function monitoring:
• To monitor for non-convulsive seizures/non-convulsive status epileptics.
• To monitor the effect of sedative/anaesthesia in patients in medically induced coma
(for treatment of refractory seizures or increased intracranial pressure).
If a patient with epilepsy is being considered for respective surgery, it is often
necessary to localize the focus (source) of the epileptic brain activity with a resolution greater
than what is provided by scalp EEG. This is because the cerebrospinal fluid, skull and scalp
smear the electrical potentials recorded by scalp EEG. In these cases, neurosurgeons typically
implant strips and grids of electrodes (or penetrating depth electrodes) under the durra mater,
through either a craniotomy or a burr hole. The recording of these signals is referred to as

Sasi Institute of Technology and Engineering Page 4


A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

electrocorticography (ECG), subdural EEG (SDEEG) or intracranial EEG (ICEEG)--all terms


for the same thing. The signal recorded from ECOG is on a different scale of activity than the
brain activity recorded from scalp EEG. Low voltage, high frequency components that cannot
be seen easily (or at all) in scalp EEG can be seen clearly in ECOG. Further, smaller
electrodes (which cover a smaller parcel of brain surface) allow even lower voltage, faster
components of brain activity to be seen. Some clinical sites record from penetrating
microelectrodes.
1.1.3 Research Use
EEG, and its derivative, ERPs, are used extensively in neuroscience, cognitive
science, cognitive psychology, and psycho physiological research. Many techniques used in
research contexts are not standardized sufficiently to be used in the clinical context. A
different method to study brain function is functional magnetic resonance imaging (FMRI).
Some benefits of EEG compared to FMRI:
• Hardware costs are significantly lower for EEG sensors versus an FMRI machine.
• EEG sensors can be deployed into a wider variety of environments than can a bulky,
immobile FMRI machine.
• EEG enables higher temporal resolution, on the order of milliseconds, rather than
seconds.
• EEG is relatively tolerant of subject movement versus an FMRI (where the subject
must remain completely still).
• EEG is silent, which allows for better study of the responses to auditory stimuli.
• EEG does not aggravate claustrophobia.

• Significantly lower spatial resolution.


• ERP studies require relatively simple paradigms, compared with block-design FMRI
studies.
EEG recordings have been successfully obtained simultaneously with FMRI scans,
though successful simultaneous recording requires that several technical issues be overcome,
such as the presence of ballista cardio graphic artifact, MRI pulse artifact and the induction of
electrical currents in EEG wires that move within the strong magnetic fields of the MRI. EEG
also has some characteristics that compare favourably with behavioural testing:

Sasi Institute of Technology and Engineering Page 5


A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

• EEG can detect covert processing (i.e., processing that does not require a response)
• EEG can be used in subjects who are incapable of making a motor response
• Some ERP components can be detected even when the subject is not attending to the
stimuli
• As compared with other reaction time paradigms, ERPs can elucidate stages of
processing (rather than just the final end result).
1.1.4 EEG Recording Method
In conventional scalp EEG, the recording is obtained by placing electrodes on the
scalp with a conductive gel or paste, usually after preparing the scalp area by light abrasion to
reduce impedance due to dead skin cells. Many systems typically use electrodes, each of
which is attached to an individual wire. Some systems use caps or nets into which electrodes
are embedded; this is particularly common when high-density arrays of electrodes are
needed. Electrode locations and names are specified by the International 10–20 system for
most clinical and research applications (except when high-density arrays are used). This
system ensures that the naming of electrodes is consistent across laboratories. In most clinical
applications, 19 recording electrodes (plus ground and system reference) are used.
A smaller number of electrodes are typically used when recording EEG from
neonates. Additional electrodes can be added to the standard set-up when a clinical or
research application demands increased spatial resolution for a particular area of the brain.
High-density arrays (typically via cap or net) can contain up to 256 electrodes more-or-less
evenly spaced around the scalp. Each electrode is connected to one input of a differential
amplifier (one amplifier per pair of electrodes); a common system reference electrode is
connected to the other input of each differential amplifier. These amplifiers amplify the
voltage between the active electrode and the reference (typically 1,000–100,000 times, or 60–
100 dB of voltage gain). In analogue EEG, the signal is then filtered (next paragraph), and the
EEG signal is output as the deflection of pens as paper passes underneath.
Most EEG systems these days, however, are digital, and the amplified signal is
digitized via an analogue-to-digital converter, after being passed through an anti-aliasing
filter. Analogue-to-digital sampling typically occurs at 256–512 Hz in clinical scalp EEG;
sampling rates of up to 20 kHz are used in some research applications. During the recording,
a series of activation procedures may be used.

Sasi Institute of Technology and Engineering Page 6


A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

These procedures may induce normal or abnormal EEG activity that might not
otherwise be seen. These procedures include hyperventilation, phonic stimulation (with a
strobe light), eye closure, mental activity, sleep and sleep deprivation. During (inpatient)
epilepsy monitoring, a patient's typical seizure medications may be withdrawn. The digital
EEG signal is stored electronically and can be filtered for display. Typical settings for the
high-pass filter and a low-pass filter are 0.5-1 Hz and 35–70 Hz, respectively. The high-pass
filter typically filters out slow artifact, such as electro galvanic signals and movement artifact,
whereas the low-pass filter filters out high-frequency artifacts, such as electromyography
signals.
An additional notch filter is typically used to remove artifact caused by electrical
power lines (60 Hz in the United States and 50 Hz in many other countries). As part of an
evaluation for epilepsy surgery, it may be necessary to insert electrodes near the surface of
the brain, under the surface of the durra mater. This is accomplished via burr hole or
craniotomy. This is referred to variously as "electrocorticography (ECOG)", "intracranial
EEG (I-EEG)" or "subdural EEG (SD-EEG)". Depth electrodes may also be placed into brain
structures, such as the amygdale or hippocampus, structures, which are common epileptic
foci and may not be "seen" clearly by scalp EEG.
The electrocorticographic signal is processed in the same manner as digital scalp EEG
(above), with a couple of caveats. ECOG is typically recorded at higher sampling rates than
scalp EEG because of the requirements of Nyquist theorem—the subdural signal is composed
of a higher predominance of higher frequency components. Also, many of the artifacts that
affect scalp EEG do not impact ECOG, and therefore display filtering are often not needed.
A typical adult human EEG signal is about 10µV to 100 µV in amplitude when
measured from the scalp and is about 10–20 mV when measured from subdural electrodes.
Since an EEG voltage signal represents a difference between the voltages at two electrodes,
the display of the EEG for the reading encephalographic may be set up in one of several
ways. The representation of the EEG channels is referred to as a montage.
 Bipolar montage:

Each channel (i.e., waveform) represents the difference between two adjacent
electrodes. The entire montage consists of a series of these channels. For example, the
channel "Fp1-F3" represents the difference in voltage between the Fp1 electrode and the F3

Sasi Institute of Technology and Engineering Page 7


A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

electrode. The next channel in the montage, "F3-C3," represents the voltage difference
between F3 and C3, and so on through the entire array of electrodes.
 Referential montage:

Each channel represents the difference between a certain electrode and a designated
reference electrode. There is no standard position for this reference; it is, however, at a
different position than the "recording" electrodes. Midline positions are often used because
they do not amplify the signal in one hemisphere vs. the other. Another popular reference is
"linked ears," which is a physical or mathematical average of electrodes attached to both
earlobes and mastoids.
 Laplacian montage:

Each channel represents the difference between an electrode and a weighted average
of the surrounding electrodes. When analogue (paper) EEGs is used, the technologist
switches between montages during the recording in order to highlight or better characterize
certain features of the EEG. With digital EEG, all signals are typically digitized and stored in
a particular (usually referential) montage; since any montage can be constructed
mathematically from any other, the EEG can be viewed by the electroencephalographic in
any display montage that is desired. The EEG is read by a neurologist, optimally one who has
specific training in the interpretation of EEGs. This is done by visual inspection of the
waveforms, called graph elements. The use of computer signal processing of the EEG—so-
called quantitative EEG—is somewhat controversial when used for clinical purposes
(although there are many research uses)
1.1.5 Normal Activity
The EEG is typically described in terms of (1) rhythmic activity and (2) transients.
The rhythmic activity is divided into bands by frequency. To some degree, these frequency
bands are a matter of nomenclature (i.e., any rhythmic activity between 8–12 Hz can be
described as "alpha"), but these designations arose because rhythmic activity within a certain
frequency range was noted to have a certain distribution over the scalp or a certain biological
significance. Frequency bands are usually extracted using spectral methods (for instance
Welch) as implemented for instance in freely available EEG software such as EEGLAB.

Sasi Institute of Technology and Engineering Page 8


A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

Most of the cerebral signal observed in the scalp EEG falls in the range of 1–20 Hz
(activity below or above this range is likely to be artifactual, under standard clinical recording
techniques).
The task of signal segments classification forms another problem that can be solved
by neural networks in many cases. The paper presents wavelet signal features classification
by self-organizing neural networks and it mentions a possible compression of signal features
as well. The method presented in the paper is applied for an EEG signal analysis and its
segments classification into the proposed number of classes.
EEG Signal Preprocessing
Information content of EEG signals is essential for detection of many problems of
the brain and in connection with analysis of magnetic resonance images it forms one of the
most complex diagnostic tools. To extract the most important properties of EEG
observations it is necessary to use efficient mathematical toolsDigital filters can be used in
the initial stage of EEG data processing to remove power frequency from the observed
signal and to reduce its undesirable frequency components. The basic principle and
application of wavelet transform is described in the first part of the contribution resulting in
the given signal wavelet feature extraction and feature vector definition.
1.1.6 Comparison Table
Table 1.1: Comparison of EEG bands
Type Frequency Location Normally Pathologically
(Hz)
Delta up to 4 frontally in • adults slow wave sleep •sub cortical lesions
adults, • in babies • diffuse lesions
posterior in •Has been found during •metabolic
children; high some continuous attention encephalopathy
amplitude tasks. hydrocephalus
waves • deep midline
lesions

Sasi Institute of Technology and Engineering Page 9


A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

Theta 4 – 7 Hz Found in • young children •focal sub cortical


locations not • drowsiness or arousal in lesions
related to task older children and adults •metabolicencephal
at hand • idling opathy
•Associated with inhibition •deep midline
of elicited responses (has disorders
been found to spike in •some instances of
situations where a person hydrocephalus
is actively trying to repress
a response or action).
Alpha 8 – 12 Hz Posterior •relaxed/reflecting • coma
regions of • closing the eyes
head, both •Also associated with
sides, higher in inhibition control,
amplitude on seemingly with the
dominant side. purpose of timing
Central sites inhibitory activity in
(c3-c4) at rest. different locations across
the brain.
Beta 12 – 30 Hz both sides, • alert/working
symmetrical • active, busy or anxious
distribution, thinking, active
most evident concentration
frontally; low
amplitude
waves
Gamma 30 – 100 + Soma to • Displays during cross- • A decrease in
sensory cortex modal sensory processing gamma band
(perception that combines activity may be
two different senses, such associated with
as sound and sight) cognitive decline,
• Also is shown during especially when
short term memory related the theta
matching of recognized band; however, this
objects, sounds, or tactile has not been proven
sensations for use as a clinical
diagnostic
measurement yet
1.1.7 Wave Patterns
Delta is the frequency range up to 4 Hz. It tends to be the highest in amplitude and the
slowest waves. It is seen normally in adults in slow wave sleep. It is also seen normally in
babies. It may occur focally with sub cortical lesions and in general distribution with diffuse
lesions, metabolic encephalopathy hydrocephalus or deep midline lesions as shown in fig1.1.

Sasi Institute of Technology and Engineering Page 10


A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

It is usually most prominent frontally in adults (e.g. FIRDA - Frontal Intermittent Rhythmic
Delta) and posterior in children (e.g. OIRDA - Occipital Intermittent Rhythmic Delta).

Figure 1.1: Delta waves


Theta is the frequency range from 4 Hz to 7 Hz. Theta is seen normally in young
children. It may be seen in drowsiness or arousal in older children and adults; it can also be
seen in meditation. Excess theta for age represents abnormal activity. It can be seen as a focal
disturbance in focal sub cortical lesions; it can be seen in generalized distribution in diffuse
disorder or metabolic encephalopathy or deep midline disorders or some instances of
hydrocephalus as shown in fig 1.2. On the contrary this range has been associated with
reports of relaxed, meditative, and creative states.

Figure 1.2: Theta waves


Alpha is the frequency range from 8 Hz to 12 Hz. Hans Berger named the first
rhythmic EEG activity he saw, the "alpha wave." This is activity in the 8–12 Hz range seen in
the posterior regions of the head on both sides, being higher in amplitude on the dominant
side. It is brought out by closing the eyes and by relaxation. It was noted to attenuate with eye
opening or mental exertion. This activity is now referred to as "posterior basic rhythm," the
"posterior dominant rhythm" or the "posterior alpha rhythm." The posterior basic rhythm is
actually slower than 8 Hz in young children (therefore technically in the theta range). In
addition to the posterior basic rhythm, there are two other normal alpha rhythms that are

Sasi Institute of Technology and Engineering Page 11


A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

typically discussed as shown in fig 1.3: the mu rhythm and a temporal "third rhythm". Alpha
can be abnormal; for example, an EEG that has diffuse alpha occurring in coma and is not
responsive to external stimuli is referred to as "alpha coma".

Figure 1.3: Alpha waves


Mu rhythm is alpha-range activity that is seen over the sensor motor cortex as shown
in fig 1.4. It characteristically attenuates with movement of the contra lateral arm (or mental
imagery of movement of the contra lateral arm).

Figure 1.4: sensor motor rhythm


Beta is the frequency range from 12 Hz to about 30 Hz. It is seen usually on both
sides in symmetrical distribution and is most evident frontally. Beta activity is closely linked
to motor behaviour and is generally attenuated during active movements as shown in fig 1.5.
Low amplitude beta with multiple and varying frequencies is often associated with active,
busy or anxious thinking and active concentration. Rhythmic beta with a dominant set of
frequencies is associated with various pathologies and drug effects, especially
benzodiazepines. It is the dominant rhythm in patients who are alert or anxious or who have
their eyes open.

Sasi Institute of Technology and Engineering Page 12


A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

Figure 1.5: Beta waves


Gamma is the frequency range approximately 30–100 Hz. Gamma rhythms are
thought to represent binding of different populations of neurons together into a network for
the purpose of carrying out a certain cognitive or motor function is shown in 1.6 given below.

Figure 1.6: Gamma waves


1.2 ARTIFACTS
1.2.1 Biological Artifacts
Electrical signals detected along the scalp by an EEG, but that originate from non-
cerebral origin are called artifacts. EEG data is almost always contaminated by such artifacts.
The amplitude of artifacts can be quite large relative to the size of amplitude of the cortical
signals of interest. This is one of the reasons why it takes considerable experience to correctly
interpret EEGs clinically. Some of the most common types of biological artifacts include:
• Eye-induced artifacts (includes eye blinks, eye movements and extra-ocular muscle
activity).
• EKG (cardiac) artifacts.
• EMG (muscle activation)-induced artifacts
• Gloss kinetic artifacts.
The most prominent eye-induced artifacts are caused by the potential difference
between the cornea and retina, which is quite large compared to cerebral potentials. When the

Sasi Institute of Technology and Engineering Page 13


A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

eyes and eyelids are completely still, this cornea-retinal dipole does not affect EEG.
However, blinks occur several times per minute, the eyes movements occur several times per
second. Eyelid movements, occurring mostly during blinking or vertical eye movements,
elicit a large potential seen mostly in the difference between the Electrooculography (EOG)
channels above and below the eyes.
An established explanation of this potential regards the eyelids as sliding electrodes
that short-circuit the positively charged cornea to the extra-ocular skin. Rotation of the
eyeballs, and consequently of the cornea-retinal dipole, increases the potential in electrodes
towards which the eyes are rotated, and decrease the potentials in the opposing electrodes.
Eye movements called saccades also generate transient electromyography potentials, known
as saccadic spike potentials (SPs).
The spectrum of these SPs overlaps the gamma-band (see Gamma wave), and
seriously confounds analysis of induced gamma-band responses, requiring tailored artifact
correction approaches. Purposeful or reflexive eye blinking also generates electromyography
potentials, but more importantly there is reflexive movement of the eyeball during blinking
that gives a characteristic artifactual appearance of the EEG (see Bell's phenomenon).
Eyelid fluttering artifacts of a characteristic type were previously called Kappa
rhythm (or Kappa waves). It is usually seen in the prefrontal leads, that is, just over the eyes.
Sometimes they are seen with mental activity. They are usually in the Theta (4–7 Hz) or
Alpha (8–13 Hz) range. They were named because they were believed to originate from the
brain. Later study revealed they were generated by rapid fluttering of the eyelids, sometimes
so minute that it was difficult to see. They are in fact noise in the EEG reading, and should
not technically be called a rhythm or wave. Therefore, current usage in
electroencephalography refers to the phenomenon as an eyelid fluttering artifact, rather than a
Kappa rhythm (or wave) as shown in the given table.
Brain electrical activity recordings byelectroencephalography (EEG) are often
contaminated with signal artifacts. Procedures for automated removal of EEGartifacts are
frequently sought for clinical diagnostics and braincomputer interface (BCI) applications. In
recent years, a combination of independent component analysis (ICA) anddiscrete wavelet
transform (DWT) has been introduced asstandard technique for EEG artifact removal.
However, in performing the wavelet-ICA procedure.

Sasi Institute of Technology and Engineering Page 14


A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

Table 1.2 current brain sensing technologies and their primary disadvantages for
HCI research.
Brain Sensing Technology Primary Disadvantage
Electrocardiogram (ECG) Highly invasive, surgery
Magneto-encephalography (MEG) Extremely expensive
Computed Tomography (CT) Only anatomical data
Single Photon Emission Computerized Radiation exposure
Tomography (SPECT)
Positron Emission Tomography (PET) Radiation exposure
Magnetic Resonance Imaging (MRI) Only anatomical data
Functional Magnetic Resonance Imaging Extremely expensive
(fMRI)
Event-Related Optical Signal / Functional Still in infancy, currently expensive
Near-Infrared (EROS/fNIR)

EEG signal analysis is such an important thing for disease analysis and brain–
computer analysis. Using Electroencephalography (EEG) monitoring the state of the user’s
brain functioning and treatment for any psychological disorder, where the difficulty in
learning and comprehending the arithmetic exists and it could allow for analysis disease the
user to train the corresponding brain. In this paper, we proposed a method for EEG signal
processing includes signal de-noising, segmentation of de-noise signal using PCM and signal
segments feature extraction done using wavelet as an alternative to the commonly used
discrete Fourier transform (DFT).These feature classified using support vector machine
classifier, Using the Matlab software proposed method accompanied.
Some of these artifacts can be useful in various applications. The EOG signals, for
instance, can be used to detect and track eye-movements, which are very important in
polysomnography, and is also in conventional EEG for assessing possible changes in
alertness, drowsiness or sleep.
EKG artifacts are quite common and can be mistaken for spike activity. Because of
this, modern EEG acquisition commonly includes a one-channel EKG from the extremities.
This also allows the EEG to identify cardiac arrhythmias that are an important differential

Sasi Institute of Technology and Engineering Page 15


A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

diagnosis to syncope or other episodic/attack disorders. Gloss kinetic artifacts are caused by
the potential difference between the base and the tip of the tongue. Minor tongue movements
can contaminate the EEG, especially in parkinsonian and tremor disorders. In this paper, we
proposed a method for EEG signal processing includes signal de-noising, segmentation of de-
noise signal using PCM and signal segments feature extraction done using wavelet as an
alternative to the commonly used discrete Fourier transform (DFT).
1.2.2 Environmental Artifacts
In addition to artifacts generated by the body, many artifacts originate from outside
the body. Movement by the patient, or even just settling of the electrodes, may cause
electrode pops, spikes originating from a momentary change in the impedance of a given
electrode. Poor grounding of the EEG electrodes can cause significant 50 or 60 Hz artifact,
depending on the local power system's frequency. A third source of possible interference can
be the presence of an IV drip; such devices can cause rhythmic, fast, low-voltage bursts,
which may be confused for spikes.
The EEG is measured and sampled while the user performs different mental tasks;
e.g., imagination of moving the left or right hand. In each BCI system, particular
preprocessing and feature extraction methods are applied to EEG samples of certain length. It
is then possible to detect task-specific patterns from EEG samples with a certain level of
accuracy.From the physiological point of view, our cortical organization and dynamics are
individual and reflect our personal life experience [2], so EEG signal variations are individual
and have fundamental differences in differentsubjects that make it impossible to design a
universal EEGbased BCI system.
For brain signal acquisition various methods used such as electroencephalography
(EEG), Functional Magnetic Resonance Imaging (FMRI), Near Infra-Red Spectroscopy
(NIRS) and Magneto encephalography (MEG). From this method EEG is wielding used
signal acquisition method because of high temporal resolution and safe for use [1]..That EEG
signal processing important for proper analysis disease. Signal processing of EEG is
fundamental for analysis of brain activity and diagnosis of normality or abnormality of signal
that is important for analysis of any disease. In this paper devote on EEG signal processing
that followed by signal de-noising, segmentation of de-noise signal using "Principal

Sasi Institute of Technology and Engineering Page 16


A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

Component Analysis (PCA)" it forms the feature vector. The paper devoted on EEG signal
processing,
EEG Data

Signal Denoising

Signal Segmentation

Figure 1.7: EEG Signal Processing


During EEG recording many of other influence introduce noise which called as
artifact. These artifacts come from patient body or instrument, as an example eyes movement,
the heart, muscles and line power. Before processing EEG, removal of this artifact is primary
task and it form fundamental step for EEG signal processing.
EEG signal de-noising can be done by both wavelet and time domain method. Here
we used Lowpass filter for de-noising. Most of time EEG signal contain neural information
below 100 Hz so it’s beneficial to use low pass filter.

Sasi Institute of Technology and Engineering Page 17


A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

Chapter 2
LITERATURE SURVEY
Electroencephalography (EEG) is an electrophysiological monitoring method to
record electrical activity of the brain. It is typically non invasive, with the electrodes placed
along the scalp, although invasive electrodes are sometimes used such as in
electrocorticography. EEG measures voltage fluctuations resulting from ionic current within
the neurons of the brain. In clinical contexts, EEG refers to the recording of the brain's
spontaneous electrical activity over a period of time, as recorded from multiple electrodes
placed on the scalp [5].
In the process of gathering brain electrical signals, inevitably, various disturbances
like the power frequency disturbance, breath disturbance, the scalp electrode's vibration and
so on can produce harmful noise signals in the brain electrical signals. These noise signal not
only submerges the characteristics of brain electrical signals, but also bring difficulties to
doctor's analysis’s, diagnosis’s even leading to a misdiagnoses, and the further analysis’s of
brain electrical signals like neural network classification, wavelet analysis’s, the analysis of
non-linear dynamics methods [5].
Various denoising techniques have been implemented for removal of the artifacts
from the EEG signals. Some of the techniques that can be used for the noise removal are ICA
denoising PCA method of denoising, Wavelet based denoising, and Wavelet packet based
denoising and so on. All the above methods can be implemented for the denoising of the EEG
signals.
E.Tamil [1] proposed that Principal component analysis (PCA) involves a
mathematical procedure that transforms a number of (possibly) correlated variables into a
(smaller) number of uncorrelated variables called principal components. The first principal
component accounts for as much of the variability in the data as possible, and each
succeeding component accounts for as much of the remaining variability as possible.
Principal components are guaranteed to be independent only if the data set is jointly normally
distributed. PCA is sensitive to the relative scaling of the original variables. Depending on the
field of application, it is also named the discrete Karhunen–Loève transform (KLT), the
Hostelling transform or proper orthogonal decomposition (POD).

Sasi Institute of Technology and Engineering Page 18


A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

G.Molina [3] proposed the mathematical technique used in PCA is called Eigen
analysis and to solve for the Eigen values and eigenvectors of a square symmetric matrix with
sums of squares and cross products. The eigenvector associated with the largest Eigen value
has the same direction as the first principal component. The eigenvector associated with the
second largest Eigen value determines the direction of the second principal component. The
sum of the Eigen values equals the trace of the square matrix and the maximum number of
eigenvectors equals the number of rows (or columns) of this matrix. PCA is sensitive to the
scaling of the variables. If we have two variables and they have the same sample variables
and are positively correlated, then the PCA will tend to rotate by 450 and the loadings for the
two variables with respect to the principal components will be equal.
A.Akrami [4] proposed that PCA is mathematically defined as an orthogonal linear
transformation that transforms the data to a new coordinate system such that the greatest
variance by any projection of the data comes to lie on the first coordinate (called the first
principal component), the second greatest variance on the second coordinate, and so on.
AbdelhamidSubasi [6] proposed another important approach for denoising the EEG
signal is the ICA method of denoising. An ICA based denoising method has been developed
by Hyvarinen and his Co-workers. The basic motivation behind this method is that the ICA
components of many signals are often very sparse so that one can remove noises in the ICA
domain. The ICA model assumes a linear mixing model x= AS, where x is a random vector
of observed signals, A is a square matrix of constant parameters, and s is a random vector of
statistically independent source signals. Each component of s is a source signal. Note that the
restriction of A being square matrix is not theoretically necessary and is imposed only to
simplify the presentation. Also in the mixing model we do not assume any distributions for
the independent components.
Subasi.A [8] proposed that ICA usually carries all the information in a single
component and most of the times this component carries non- artifactual information which
may lead to information loss. Also ICA performance depends on the dataset size. Another
limitation which arose in this method is that the signals can be analysed only in time domain
not in the frequency domain as the artifacts in EEG have a typical frequency range and are
overlapped with the spectrum of the EEG data this becomes one of the disadvantage of this
method.

Sasi Institute of Technology and Engineering Page 19


A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

Ubeyli.E.D and Wang.X [9] proposed the term ‘wavelet’ refers to an oscillatory
vanishing wave with time-limited extend, which has the ability to describe the time-
frequency plane, with atoms of different time supports. Generally, wavelets are purposefully
crafted to have specific properties that make them useful for signal processing. They
represent a suitable tool for the analysis of non-stationary or transient phenomena.
Spa.J.Le.D.
Tan [2] says that in wavelet denoising we decompose the signals in to high frequency
components and low frequency components using the threshholding method and apply
wavelet transform to the low frequency components. The two thresholding methods available
are hard threshholding and soft threshholding. And then we select the best wavelet from the
wavelet families which can best decompose the noisy signal and again we reconstruct the
signals.
Hence various methods have been studied for demising of the EEG signal. We know
that a denoised signal has high PSNR(peak signal to noise ratio),SNR(signal to noise ratio)
and low MSE(mean square error).By considering above performance measures, we came to
know that they deals only with time domain, in which noise cannot be removed completely.
So we proposed a method called as “DUAL TRACE COMPLEXE WAVELET”. The main
advantage of this method is that it removes the noise completely from the EEG signal and
hence we proposed this method.
The state of brain is in a continuous change, with EEG having different spectral
properties depending on the behavioral states (e.g., sleep, awake …) and cognitive tasks
being undertaken. Numerous studies have demonstrated correlation between EEG signals and
actual or imagined movement, Many different features are used in EEG signal processing
applications. Based on previous studies [8], features extracted in frequency domain are one of
the best to recognize the mental tasks based on EEG signals.
EEG signal is nonstationary that means its spectrum changes with time; but we
assume that the spectral characteristics of the EEG are changing continuously and slowly.
Such a signal can be approximated as piecewise stationary, a sequence of independent
stationary signal segments. The field of spectral analysis has been dominated by use of the
Fourier transform. The Fourier functions do not adequately represent nonstationary signals.

Sasi Institute of Technology and Engineering Page 20


A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

EEG is usually treated as a realization of an underlying neurophysiological process


whose spectrum we want to estimate. The standard approach to estimating the spectrum is to
compute the periodograms, which are simply the square of the magnitude of the Fourier
coefficients [9]. For segmentation of the EEG signal, it is necessary to use a moving time
window that is small enough to approximate the stationarity of the underlying process.
However, if the time window is too small, the resolution in the frequency domain diminishes
and components of the signal that oscillate at similar frequencies will be indistinguishable.
Therefore, a tradeoff between stationarity and frequency resolution is inevitable.
During whether classification or clustering, frequency features extracted from the
power spectrum should be discriminated. Many different classification methods are used for
recognition of EEG patterns associated to mental tasks. In this study, we use linear clustering
method, neural networks, and HMM classifiers to discriminate EEG data into three different
classes of mental tasks: 1) Baseline (relaxation) class, 2) Right hand movement imagery
class, 3) Left hand movement imagery class.
A Mental Tasks
The goal of this BCI system is segregation of movement imagery of left and right
hand according to the spontaneous EEG signals. The EEG signal is elicited by having the
subject, a 24 years old right-handed male, execute different mental tasks while remaining in a
totally passive state. This implies that no overt movements are made during the performance
of the tasks.
The subject should sit comfortably, and execute the motor imagery tasks upon
request. Electrodes are placed according to 10-20 standard. Although recording of EEG is
from 24 channels, only 3 channels will be used in each feature extraction method. Another
channel is also used to record EOG signal for removing eye blink artifacts. Data from all of
electrodes are recorded for 15-second intervals during each task and each task is repeated 10
times per session, and there are totally 2 sessions. The rest time between trials is nearly 1
minute.
There are a total of 3 tasks and each task is performed under eyes closed condition.
Since the imagination related EEG signals differ significantly from task to task, we should
have the subject do a specific imagination during different trials. So, clockwise turning the

Sasi Institute of Technology and Engineering Page 21


A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

arm around the shoulder with a constant velocity is the mental task the subject should
imagine.
Pre-processing including filtering, digitizing, artifact removing and DC level
correction are to be applied on the EEG signals. EEG is amplified and filtered with a band-
pass filter with band-width of 0.1-80 Hz and then digitized at 256 Hz in 12 bits. B Artifact
Removing
EOG artifact is one of the most important artifact sources in EEG which is due to eye
blink or eyeball movement. This artifact affects mainly the signals from the most frontal
electrodes (Fp1 and Fp2 and also other frontal electrodes: F3, F4, F7 and F8), and induces
many high and low frequencies in them, depending upon its duration and amplitude. There
are many different methods or partially simple criteria for artifact recognition [11]. Classical
methods for removing EOG artifacts can be classified into rejection methods and subtraction
methods.
CSegmentation
Because of nonstationary nature of EEG, artifact-removed EEG signals, which are
less than 15 seconds, should be segmented in smaller intervals in which they are
approximated as piecewise stationary. The processing is only applied on 11 seconds of each
trial that means we skip first and the last segments of each trial. Each 11-second trial is
segmented in 2-second pieces with 1 second (50%) overlap. DFeatures

Features extracted in this BCI system, are logarithmic power of different frequency
bands of EEG which are extracted from various combinations of channels.There is a main
assumption in all power estimation methods which implies the signal is the summation of
some non-correlated frequency harmonics. Therefore, power density of each harmonic is to
be calculated for power estimation of the main signal. We have estimated the power spectrum
by periodogram method in which, Fourier coefficients of the signal are calculated by a 512-
point Fast Fourier Transform. Then, square of the amplitude of the Fourier coefficients are
used as the power of EEG samples.
The squared coefficients produce a new series of numbers on the frequency scale,
which comprise the raw spectrum of the original EEG segment. Finally, the numbers in the
raw spectrum are averaged together in groups defined by interesting frequency ranges, to

Sasi Institute of Technology and Engineering Page 22


A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

yield the power in the standard bands: delta, theta, alpha, beta, and gamma [13], and then
power spectral density components are transformed to dB, and normalized to the total energy
of the 2-second segment.
E Brain Sensing
The human brain is a dense network consisting of approximately 100 billion nerves
cells called neurons. Each neuron communicates with thousands of others to regulate physical
processes and produce thought. Neurons communicate either by sending electrical signals to
other neurons through physical connections or by exchanging chemicals called
neurotransmitters. Advances in brain sensing technologies enable us to observe the electrical,
chemical, or blood flow changes as the brain processes information or responds to various
stimuli.
EEG for Task Classification
Based on the results from pilot recordings with our system, we chose three
tasks, 1- Rest
In this task, our baseline, we instructed participants to relax and to try not to focus on
anything in particular. We also explicitly instructed them not to continue working on any task
that may have preceded the rest task.
2- Mental Arithmetic
In this task, participants performed mental multiplication of a single digit number by a
three digit number, such as 7 × 836. We chose the complexity of the problems so that it was
not so difficult as to be discouraging, but also so that it would take most participants more
than the allotted time to complete it. We instructed participants to double check their answers
if they finished before the time expired. This ensured that they were performing the intended
task as well as they could throughout the task period.
3- Mental Rotation
In this task, participants imagined specific objects, such as a peacock, in as much
detail as possible and rotating in space. The specific details of the object were left to the
participant.
In order to classify the signals measured from our EEG, we performed some basic
signal processing to transform the time series data into a time independent data set. We then

Sasi Institute of Technology and Engineering Page 23


A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

computed a set of base features that we mathematically combined to generate a much larger
set of features.
Next, we used a feature selection process to prune the feature set, keeping only those
that added the most useful information to the classifier and to prevent over-fitting. Our
feature generation and selection process was similar to that used by Fogarty et al. in their
work on modeling task engagement to predict interruptibility [9]. We used these features to
train a Bayesian Network and perform the classification.

Sasi Institute of Technology and Engineering Page 24


A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

Chapter 3
WAVELETS & EEG
3.1 Wavelet Transform
Wavelet Transform is a time scale analysis method and has the capacity of
representing local characteristics in the time and frequency domain. In the low frequency, it
has a lower time resolution and high frequency resolution, the high frequency part has the
high time resolution and low frequency resolution.it is suitable for detection of normal signal.
which contains transient anomalies.

Time-domain wavelets are simple oscillating amplitude functions of time. So are the
sine and cosine waves of Fourier analysis. However, unlike sine and cosine waves which are
precisely localized in frequency but extend infinitely in time (sines and cosines have definite
single frequencies, e.g., 40Hz, constant for all time); wavelets are relatively localized in both
time and frequency.

They have large fluctuating amplitudes during a restricted time period and are very
low amplitude or zero amplitude outside of that time range. That is, wavelets are said to be
‘‘supported’’ over a restricted domain of time if the bulk of their energy is restricted to that
time period and are said to be ‘‘compactly’’ supported if all of their energy is restricted to a
specific domain of time.

Signal processing achieved milestones by the evolution of various signal


decomposition methods due to their unique nature in processing signals . Several wavelet
transform namely Discrete Fourier Transform ,Discreate Wavelet Transform ,Empirical mode
decomposition, signal valued decomposition and their variants were widely used for seizure
detection and prediction applications and projected their efficiency in that applications.

Although many other time-frequency feature engineering approaches prevailing for


signal processing such as EMD, SVD, ICA, and PCA, DWT based wavelet feature analysis is
identified as effective for time-frequency domain analysis due to its multiscale approximation
feature.Another highlight of DWT based feature engineering is that DWT is used for both
signal noise reduction as pre-processing and feature extraction . According to Lina Wang et
al. multiresolution analysis of feature engineering produced better EEG signal processing

Sasi Institute of Technology and Engineering Page 25


A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

results. Wavelet analysis as reported by Yatindra et al. Wavelet domain feature engineering
using DWT and its variants are used in combination with many machine learning classifiers
for epileptic seizure detection.

Many classification works were reported in the recent days using DWT but all the
works are not ideal for analysis, as the datasets used is not clear. The few works mentioned
here were used novel approaches for classification after wavelet analysis. The multiresolution
analysis using DWT is dominating now in EEG signal feature engineering for epileptic
seizure detection in combination with diversified classifiers. It is also noticed from the recent
works that the SVM and FFNN classifiers are used most with DWT as feature engineering
tool. Multiresolution analysis of signal processing and feature engineering is enhancing the
classification accuracy.

Another highlight of DWT based feature engineering is that DWT is used for both
signal noise reduction as pre-processing and feature extraction . According to Lina Wang et
al. multiresolution analysis of feature engineering produced better EEG signal processing
results. Wavelet analysis as reported by Yatindra et al.

3.1.1 Time Domain Features


Time domain features are amplitude related statistical features such as energy, power,
and mean, and variability, regularity are tested with variance, coefficient of variation and
total variation. In time domain, the epileptic seizure events are detected by analyzing
discrete time sequences of EEG segments. These kinds of analysis can be performed
through histograms of the EEG time segments.

A simple time-domain seizure detection method presented by Runarsson and


Sigurdsson and suggesting for estimating the histograms for two variables: the amplitude
difference and time gap between peak values as well as minima, as the estimated values for
the histogram bins the features for classification of an EEG segments as a seizure or non-
seizure. Time domain approach for feature engineering only provides spatial features, but
temporal information is missing. In the event of epileptic seizures, signal might have
uncertain oscillations, thus the frequency component is required to classify seizures.

Sasi Institute of Technology and Engineering Page 26


A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

3.1.2 Frequency Domain Features

Frequency is one of the prime components used to measure the occurrence of the
events at precise time. As EEG is non-stationary, different frequency bands are used to
locate the events. When an EEG signal is represented by its frequency is crucial for the
analysis of signals through wavelet transform. Based on the biomedical signal to be
analyzed, the mother wavelet is chosen, based on the classification accuracy and
computational time obtained in the experiment , it was found the best wavelet transform for
analysis of EEG signal.

For EEG seizure detection Fourier transform magnitudes are being used as frequency
domain features. Amjed S Al- Fahoum et al. used frequency domain features for comparing
the performance of classification for epileptic seizure detection with features in wavelet
domain frequency moment.As EEG signals are non-linear and non-stationary, there is a
difficulty to characterize different activities of EEG signals with certain mathematical
models. In order to address this issue, Acharya et al. proposed a method for the detection of
normal, pre-ictal, and ictal conditions from recorded EEG signals.

3.1.3 Wavelet (Time-Frequency) Domain Features


Time and frequency domain analysis of features have their own limitations in dealing
with EEG signals. Even if the time domain analysis provides spatial feature components,
the frequency content information is missing for EEG analysis. Frequency domain method
may use temporal information for extracting features, but only after windowing the
function. These limitations can be resolved by using time-frequency domain analysis.
For automatic detection of epileptic seizures that three levels DWT with Db4 wavelet
efficiently performs three-class classification using multiclass sparse extreme learning
machine. Amjed S et al. had performed a detailed study on the feature engineering with
frequency and time-frequency domain methods for EEG classification. Widely used feature
extraction methods such as TFD, FFT, EM, DWT and ARM are taken into consideration
and concluded that each method has its own advantages and highly depends on the signal
to be analyzed for the application.

Sasi Institute of Technology and Engineering Page 27


A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

3.1.4 Wavelet Family


Wavelet transform will be an effective time–frequency analysis tool for analyzing the
ephemeral nature of EEG signals as it unifies different approaches towards bio-medical
classifications. Wavelet features are playing major role in evaluating transient events in
biological signals. Variety of wavelet families is available for signal characterization. Few
widely used mother wavelet families for EEG classifications include Harr, Daubechies,
Symlet and Coiflet and etc.

Selection of suitable wavelet is crucial for the analysis of signals through wavelet
transform. Based on the biomedical signal to be analyzed, the mother wavelet is chosen
based on the classification accuracy and computational time obtained in the experiment, it
was found that Coiflet of order 1(Coif1) is the best wavelet family for analysis of EEG
signal as the support width of the mother wavelet function resembles that of the EEG
signal and also has a compact filter length, thus reducing the processing time.

3.2 Wavelet Analysis


Wavelet analysis refers to a growing class of signal processing techniques and
transforms that use wavelets and related functions called wavelet packets to efficiently
measure and manipulate such non-stationary signals. These detail functions can isolate all
scales of waveform structure, from the largest to the smallest pattern of variation in time and
space that is available in the neuroelectric data set. Consequently, wavelet analysis provides
flexible control over the resolution with which neuro electric components and events can be
localized in time, space, and scale.

3.3 Wavelet Decomposition


The output of Wavelet transformed Signals are having two Co-Efficient called as
Approximate Coefficient and Detail Co-Efficient. The Approximation Co-Efficient
corresponds to Low frequency Component and the Detail Co-Efficient corresponds to High
frequency Component of a Wavelet. This process is called Signal Decomposition.

At the beginning of a DWT computation, a neuro-electric waveform like EEG


containing n samples is run through the high and low pass filters. The output each filter is a
series of n wavelet coefficients. Every other coefficient is of discarded from the series,

Sasi Institute of Technology and Engineering Page 28


A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

leaving n/2 coefficients for each filter output. This process of discarding alternate coefficients
is known as down sampling and is indicated in the figure by the downward pointing arrow.

The output of the high pass filter is the set of DWT wavelet coefficients associated
with all of the discrete wavelets at the smallest single scale available for the particular
digitized neuro-electric waveform that went into the filter. This output captures all of the high
frequency energy in the waveform. The output of the low pass filter is the set of DWT
coefficients associated with a set of companion functions called scaling factor.

In non linear approximation we keep only a few signification coefficient of a signal


and set the rest to zero. Then we reconstruct the signal using the signal coefficients. Wavelet
produces a few signification coefficients for the signals with discontinuities. Thus we obtain
better results for wavelet transform non-linear approximation when compared to Fourier
transforms.

3.4 WAVELET MULTI RESOLUTION ANALYSIS:


WMA incorporates the steps of DWT and inverse DWT. The initial DWT consists of
sequential application of low- and high-pass filters to decompose a discrete signal into
multiple wavelet components, as shown in Figure 3.1. Here, x[n]represents a channel of EEG
signal passed through a low pass filter, g[n] and a high pass filter, h[n] simultaneously. This
process is repeated until each channel of the EEG signal is decomposed into n levels of
wavelet details, i.e. D1 (t),D2 (t), ⋯ ,Dn (t) and a mother wavelet An (t).

The Wavelet transform decomposes the EEG signal to yield the approximation co-
efficients and detail co-efficients. These co-efficients were used as input to compute the
energy of features. These values enable to extract the features associated with
stimuli.Wavelets usually utilized to de-noise biomedical signals comprising orthogonal
meyer wavelet and Daubechies, ‘db8’ ‘db6’ and ‘db2’ wavelets. These are normally selected
from the shapes similar to those EEG signals.

Sasi Institute of Technology and Engineering Page 29


A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

Figure 3.1. Three-level wavelet decomposition tree.

A wavelet decomposes a signal in various multiresolution parts in accordance with a


basic function. This is known as wavelet function. The most extensively utilized signal
processing functions are filters. The filtering operations determine the resolution of the
signal, which is a calculation of detailed information in the signal. The scale is stabilized by
downsampling or sub sampling and the upsampling operations. DWT is calculated with
consecutive highpass and lowpass discrete time domain signal filtering, which is illustrated in
figure 3.1.

DWT is an eff ective way of analyzing nonstationary EEG signals. This technique


provides high-frequency resolution if the frequency is low, and high-time resolution if the
frequency is high because it uses long time windows at low frequencies and short time
windows at high frequencies. DWT decomposes a signal into subbands by filtering of the
time domain signal f using sequential high-pass and low-pass filters.

Feature extraction is a special form of dimensionality reduction. When the input data
to an algorithm is too large to be processed and it is suspected to be notoriously redundant
(much data, but not much information) then the input data will be transformed into a reduced
representation set of features (also named features vector). Transforming the input data into
the set of features is called feature extraction. If the features extracted are carefully chosen it
is expected that the features set will extract the relevant information from the input data in
order to perform the desired task using this reduced representation instead of the full size
input. The following study is devoted to the wavelet domain signal feature extraction and
comparison of results achieved.

Sasi Institute of Technology and Engineering Page 30


A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

The basic principle and application of wavelet transform is described in the first part
of the contribution resulting in the given signal wavelet feature extraction and feature vector
definition.

The frequency content of EEG signal provides useful information than time domain
representation. The wavelet transform gives us multi-resolution description of a nonstationary
signal. EEG is non-stationary signal hence wavelet is suited for EEG signals. At high
frequencies it represents a good time resolution and for low frequencies it represents better
frequency resolution. This multi-scale feature of the Wavelet allows the decomposition of a
signal into a number of scales, each scale representing a particular coarseness of the signal
under study. The procedure of multiresolution decomposition of a signal x[n] is schematically
shown in figure 3.1.

Sasi Institute of Technology and Engineering Page 31


A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

Chapter 4
METHODOLOGY
4.1 Introduction
The electroencephalogram (EEG) is the recording of electrical activity of the cerebral
cortex through electrodes, which are usually placed on the scalp. The EEG technique is
widely used for the clinical diagnosis of epilepsy and sleep disorders. Today, the EEG is also
attracting increasing interest in brain computer interface (BCI) applications. However, in
practical settings the EEG signals are often contaminated by both biological and
environmental artifacts. Biological artifacts are signals arising from non-cerebral sources in
the human body, such as cardiac, ocular or muscles activity. On the other hand,
environmental artifacts originate from outside of the human body, due to electrode movement
or interference from external devices such as power main or electric motor. Together,
biological and environmental artifacts degrade EEG signals, thereby obstructing clinical
diagnosis or BCI applications by distorting the observed power spectrum.
EEG signals are having very small amplitudes and because of that they can be easily
contaminated by noise .The noise can be electrode noise or can be generated from the body
itself. The noises in the EEG signals are called the artifacts and these artifacts are needed to
be removed from the original signal for the proper analysis of the EEG signals. Thevarious
types of noises that can occur in the signals during recordings are the electrode noise, base
line movement, EMG disturbance and so on.We need to remove these noises from the
original EEG signal for proper processing and analysis of the diseases related tobrain.
Various denoising techniques have been implemented for removal of the artifacts
from the EEG signals. Some of the techniques that can be used for the noise removal are ICA
denoising,PCA method of denoising, Wavelet based denoising, and Wavelet packet based
denoising and soon. All the above methods can be implemented for the denoising of the EEG
signals and their performance evualuation can be done by measuring the parameters like
SNR, PSNR, and MSE etcEEG recording method could be categorized into two groups:
invasive electrode and non invasive electrode.

From the signals by selecting the best wavelet to decompose the signal. Wavelet
Packet transform was used for EMAT noise suppression which decomposes the signaling

Sasi Institute of Technology and Engineering Page 32


A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

both low pass and high pass component and shown SNR improvement of 19 dB.The
wavelet based threshold method and Principal Component Analysis (PCA) based adaptive
threshold method to remove the ocular artifacts[3]. The disadvantage of PCA is the
requirement that artifacts are uncorrelated with the EEG signal. This is a stronger
requirement than the independency requirement of ICA.

A deficiency of the invasive EEG acquisition method is it usually took more than one
month for the patient to recover completely from the surgery. The advantage of this
invasive method is its high accuracy and sensitivity. The signal to noise ratio of invasive
EEG is from 10 to 100 times higher than non-invasive EEG recording method. Currently,
invasive EEG signal recording method emphasis in brain disease diagnoses. The noise
reduction technique using independent component analysis(ICA) and subspace filtering is
presented. They applied subspace filtering not to the observed raw data mixed version of
these data obtained by ICA.. Finite impulse response filters are employed whose vectors are
parameters estimated based on signal subspace extraction. ICA allows to filter independent
components. After thenoise is removed they reconstruct the enhanced independent
components to obtain clean original signal.
In general, electroencephalography (EEG) signals are used in the analysis of these
electrical discharges that result in disorders of the brain [1]. The visual detection of epileptic
seizures and the visual diagnosis of epilepsy require the scanning of long EEG recordings,
which is a very time consuming process. Since a whole visual examination is often not
possible, the automated systems based on artificial neural networks (ANNs) are used in the
analysis of EEG signals.
Although EEG signals are non-stationary signals, most epilepsy diagnosis systems are
based on the assumption that EEG signals have quasistationary characteristics in the time or
frequency domain. In order to analyze such signals, time-frequency-based approaches are the
most suitable tools, [2,4]. The discrete wavelet transform (DWT) is the most appropriate
transform method for applications with nonstationary signals like EEG signals, since it
provides both time and frequency views of the signals simultaneously.
Several studies noted that the classification accuracy of EEG signals depends entirely on the
selection of optimum statistical parameters (such as maximum, minimum, standard variation,
mean, entropy, and average power) in not only the time or frequency domain, but also in the

Sasi Institute of Technology and Engineering Page 33


A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

time-frequency domain [9–17]. Since the entropy is a nonlinear measure and quantifies the
degree of complexity in a time series, it helps to understand brain dynamics when it is used in
the analysis of EEG signals.
. In addition, using any discretization method, the data points of EEG signals can be
divided into clusters or groups, and, in this way, hidden clusters of data points may be
discovered, and therefore the analysis of the signals may become easier. To this end, EEG
signals are decomposed into frequency subbands using the DWT method, the coefficients of
frequency subbands are discretized into the desired number (K) of intervals using the EWD
and EFD methods, the entropy values of these discretized coefficients are computed with the
Shannon entropy method, and these are then used as inputs into the ANFIS in the
classification of EEG signals related to different combinations of healthy segments
Conventional methods to remove EEG artifacts employ linear filters or regressions, in
relation to the time of occurrence or the frequency range of target artifacts. However, filtering
in either the time or frequency domains incurs substantial loss of observed cerebral activity
because of the inherent spectral overlap between neurological activity and signal artifacts.
Wavelet based multiresolution analysis using a discrete wavelet transform (DWT) is shown
to be more effective in removing target artifacts, while better preserving the structure of the
true EEG signal in both time and frequency domains. On the other hand, independent
component analysis (ICA) is proven useful to isolate target artifacts into a separated
independent component (IC) using blind source separation. In recent years, artifact removal
using a combination of wavelet and ICA methods has shown promising results in practical
applications.
Applying the joint method of wavelet-ICA for artifact removal can necessitate visual
inspection of the EEG recording, or make it necessary to apply a manually-defined or
arbitrary threshold to identify and isolate the artifactual component from the EEG signals.
The defined threshold may fail to capture target artifacts close to the arbitrarily defined
boundary of the EEG signals. Additionally, using a manually-defined threshold to identify
signal artifacts may also increase the false detection rate. Furthermore, depending on the
particular dataset being assessed, the calculated thresholding value may not be appropriate to
distinguish multiple target artifacts in cases of noisy signals recorded at multiple channels, as
is frequently observed in the case for inherently noisy multi-channel EEG.

Sasi Institute of Technology and Engineering Page 34


A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

4.2 Principle Component Analysis


Normalized Principal Components that are decor related linearly and can be remixed
to reconstruct the original EEG. PCA uses the eigenvectors of the covariance matrix of the
signal to transform the data to a new coordinate system and to find the projection of the
input data with greater variances. The components of the signals are then extracted by
projecting the signal on to the eigen vectors. PCA has been shown to be an effective method
for removing ocular artifacts from EEG signals.
Disadvantages
 One disadvantage of PCA is the requirement that artifacts are uncorrelated with the
EEG signal. This is a stronger requirement than the independency requirement of
ICA.

 It has been observed that PCA cannot completely separate eye-movement artifacts,
EMG and ECG artifacts from the EEG signal, especially when they have comparable
amplitudes.

 PCA also does not necessarily decompose similar EEG features into the same
components when applied to different epochs.
4.3 Blind Source Separation
BSS techniques separate the EEG signals into components that “build” the EEG
signals. They identify the components that are attributed to artifacts and reconstruct the
EEG signal without these components. Among the BSS methods, Independent
Component Analysis (ICA) is more widely used.BSS techniques have the ability to
separate EEG signals to spatial components; specialists are then are called then to
identify the artifactual components remove them and reconstruct the signal.
Disadvantages
 The main disadvantage is that specialist should be available to recognize and
reject the artifactual components.

 This is a time consuming procedure which can lead to cerebral activity distortion
if will not be performed with great care.
4.4 Linear Filtering
Linear filtering is useful for removing artifacts located in certain frequency bands

Sasi Institute of Technology and Engineering Page 35


A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

that do not overlap with those of the neurological phenomena of interest. For example,
low-pass filtering can be used to remove EMG artifacts and high-pass filtering can be
used to remove EOG artifacts.
Advantages
 The advantage of using filtering is its simplicity.

 Also the information from the EOG signal is not needed to remove the
artifacts.
Disadvantages
 This method, however, fails when the neurological phenomenon of interest and
the EMG, ECG or EOG artifacts overlap or lie in the same frequency band.

 As a result, a simple filtering approach cannot remove EMG or EOG artifacts
without removing a portion of the neurological phenomenon.

 More specifically, since EOG artifacts generally consist of low-frequency
components, using a high-pass filter will remove most of the artifacts and for
EMG artifacts, using a low pass filter will remove some artifacts.
Uses
 Linear filtering was commonly used in early clinical studies to remove artifacts
in EEG signals.
4.5 Independent Component Analysis
ICA model describes multivariate signals in terms of a mixing of source components,
by making the general assumption that multivariate signals, x are separable into their
statistically independent and non-Gaussian source components, s. This approach has been
widely applied in EEG signal processing to separate EEG artifacts, with the requirement that
several assumptions are met:
• The multivariate signals consist of cerebral and artifactual sources that are linearly
mixed and statistically independent..
• At most one source component is Gaussian, and
• The propagation delay of artifactual sources through the scalp is negligible.
The source components are synonymous with independent components (ICs). The
relationship between a recorded signal and its source components is described by the
equation
Sasi Institute of Technology and Engineering Page 36
A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

X=As. (1)
In equation (1), A is the unknown mixing matrix which is to be estimated by using the
ICA algorithms. Then, the inverse of matrix A can be computed as the estimated un-mixing
matrix, W. Finally, the source components, s are revealed by using the equation
S=Wx. (2)
The reconstruction of source components into the multivariate signals is known as
inverse ICA, which is accomplished by multiplying the inverse of the estimated mixing
matrix, W-1with the source components.
This algorithm is highly effective at performing source separation in domains where
In case of EEG signals, the multi-channel EEG recordings are mixtures of underlying brain
and artifact signals. Volume conduction is thought to be linear and instantaneous and hence
(a) is satisfied (b) is also reasonable because sources of eye and muscle activity, line noise
and cardiac signals are not generally time locked to the sources of EEG activity.
Assumption c) Here the effective number of statistically independent signals contributing to
scalp EEG is not known but numerical simulations have confirmed that the ICA algorithm
can accurately identify the time courses of activation and scalptopographies of relatively
large and temporally independent sources from scalp recordings even in the presence of
low-level and temporally independent source activities.
4.6 Proposed Method
We proposed a hybrid method for automatic identification and removal of artifactual
components in EEG signal, without any need to apply an arbitrary threshold in identifying the
artifactual components in fig 4.2 . In brief, our hybrid method applies a combination of
wavelet-ICA to assist in classification of artifactual ICs. It consists of following steps,

1. EEG signals are decomposed into ICs using ICA.


2. Apply wavelet transform to ICs.
3. Threshold wavelet coefficients to remove artifacts.
4. Inverse wavelet transform to obtain artifact-free ICs.

Sasi Institute of Technology and Engineering Page 37


A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

Raw EEG Noise Wavelet Multi Wavelet Multi


Signal Added EEG Resolution Resolution
Signal Analysis Analysis

Denoising
Signal

Figure 4.1: Block diagram of the proposed artifacts removal system using wavelet-ICA
4.6.1 EEG Recording
The subjects are instructed to maintain a natural upright sitting position with eyes
open for up to 30 minutes. EEG signals with eye blink artifacts are recorded following
involuntary eye blink activities. The electrodes were placed as specified by the 10-20 system.
Atonal of 16 electrodes corresponding to channels FP1, FP2, F3,Fz, F4, T7, C3, Cz, C4, T8,
P3, Pz, P4, O1, Oz, and O2 were used in this study. In our procedure, the ground electrode is
seat FPz, and the reference point fixed at the left earlobe (A1).The scalp impedance of the
recording is kept below 5 kΩ. The recordings were conducted with a sampling rate of 256
Hz. A notch filter of 50 Hz (Butterworth, order 4) and band pass filter of 0.5 to 100 Hz
(Butterworth, order 8) was applied by default during the recording, whereupon the signal was
separated into five-second epochs for further processing.
4.6.2 Procedure For Wavelet Multiresolution Analysis
WMA was first applied to the EEG recording in order to exclude all but the frequency
bands of interest. Each channel of the recorded signal is decomposed by DWT to 8 levels
using a mother wavelet of db8. By default, WMA deletes details at levels D1 and D2,
corresponding to the frequency range of 32 to 128 Hz, and also the mother wavelet A8,
corresponding to the frequency range of 0 to 0.5 Hz. As such, WMA retains relevant details
of D8 to D3, corresponding to the frequency range of interest for EEG signal, i.e. 0.5 to 32
Hz. The wavelet details represent the traditional frequency bands of EEG signals defined as
delta (0.5 to 4 Hz), theta (4 to 8 Hz), alpha (8 to 16 Hz) and beta (16 to 32 Hz) bands
respectively [20]. WMA filtered most of the artifacts out of the frequency range of interest,

Sasi Institute of Technology and Engineering Page 38


A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

notable high frequency noise (>32 Hz), and linear trend movement at extremely low
frequency (<0.5 Hz).
Signal segments feature extraction forms the next step of signal segmentation
allowing combination of time-domain and frequency-domain signal features. Commonly
used spectral representation of a signal based upon its all-pole model or its discrete Fourier
transform provides the same frequency resolution over the whole window function. To allow
different resolution the wavelet transform is often used providing its very efficient alternative
allowing different levels of decomposition.
4.6.3 ICA Decomposition
In signal processing ,independent component analysis is a computational method for
separating a multivariate signal into additive subcomponents. This is done by assuming that
the subcomponents are non-gaussian signals and that they are statistically independent from
each other. ICA is a special case of blind source separation. A common example application
is the cocktail party problem of listening in on one person’s speech in a noisy room.
After preliminary filtering of the EEG signal by WMA. The number of ICs is
constrained to be less than or equal to the number of channels of the EEG signal. We selected
the matrix-pencil algorithm over alternates such as fast ICA or the Infomax algorithm due to
its superior performance in application for non-stationary signals. Additionally, the matrix-
pencil algorithm based on second-order statistics also requires less computational load than
algorithms based on higher-order statistics.
4.6.4 Wavelet Artifact Removal
Wavelet artifact removal is applied to the ICs identified by SVM as constituting
artifactual components. The ICs are again decomposed by DWT and the wavelet components
with a coefficient exceeding the universal value for wavelet denoise is deemed to be purely
artifactual, and is thus removed. The universal value, K for wavelet denoise is calculated as
KK = √2 log Nσ, (3)
Where N is the length of the data to be processed and

σ2 = median (|W (j, k)|/0.6745) (4)


Represents the magnitude of neuronal wide band signal. In equation (4), |W (j, k)|
represents the absolute value of the wavelet coefficient, with constant 0.6745 accounting for
the Gaussian noise.
Sasi Institute of Technology and Engineering Page 39
A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

In EEG recording, a powerful electronic amplifier increases several hundreds or


thousands of times the amplitude of the weak signal (less than a few micro volts) which is
generated in this place. In the past, a device called galvanometer, which has a pen attached to
its pointer, writes on the paperstrip,which moves continuously at a fixed speed past it. In the
present time, with the advent of powerful electronic computer and very high storage, we can
use A/D device to transform signal between electrode and computer. A lot of data can be
recorded and easily analyzed and printed.
One pair of electrodes usually makes up a channel. Since earlier times, it is known
that the characteristics of EEG activity change in many different situations, particularly with
the level of vigilance: alertness, rest, sleep and dreaming. The frequency of wave change can
be labelled with names such as alpha, beta, theta and delta. Particular mental tasks also alter
the pattern of the waves in different parts of the brain. A small picture to show how eye blink
contaminates EEG Signal.
The current de-noising techniques, that are based on the frequency selective filtering,
suffer from a substantial loss of the EEG data. Preventing patients from a normal work is not
considered a feasible solution. In fact, this may have major impacts in recording EEG.
Considering the problems, frequency selective filtering technique for eliminating noise from
EEG is regarded as a challenging task nowadays . An attractive substitute is the Wavelet
based filtering considering the capability of studying both frequency and time maps
simultaneously . Stationary wavelet transform (SWT) is utilized for de-noising EEG signal by
Zikov et al. Reconstructed signal in SWT technique is not a good approximation for original
EEG because artifacts associated to recorded EEG signal are considerably uncorrelated.a
discrete wavelet-based noise elimination is carried out to get rid of artifacts from EEG signal.
In de-noising physiological signals.
The task of signal segments classification forms another problem that can be solved
by neural networks in many cases. The paper presents wavelet signal features classification
by self-organizing neural networks and it mentions a possible compression of signal features
as well. The method presented in the paper is applied for an EEG signal analysis and its
segments classification into the proposed number of classes.

Sasi Institute of Technology and Engineering Page 40


A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

4.6.5 Wavelet And ICA Reconstruction


After the removal of artifacts, the remaining wavelet components are reconstructed into ICs
by inverse DWT. Finally, inverse ICA is applied to reconstruct the filtered ICs into clean
EEG signals with artifacts removed. Wavelet de-noising is efficient as it has a tendency for
preserving signal characteristics while reducing noise, this is favoured over signal frequency
domain filtering. The reason is that the threshold strategies are available which allows
reconstruction based on chosen coefficients.

Sasi Institute of Technology and Engineering Page 41


A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

Chapter 5
RESULTS
5.1 Experimental Results:
In this section, the results of cardiac artifacts removal from EEG signal using
wavelet-ICA was discussed. The EEG signal is taken from the database available in the
website http://www.physionet.org/pn6/chbmit/.

Figure 5.1: Raw EEG signal


Obtained Raw EEG signal in the figure 5.1 the can be classified into different types
depending upon amplitudes and frequencies in figure 5.2.

Sasi Institute of Technology and Engineering Page 42


A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

Figure 5.2: Raw EEG signal classification


And that raw EEG signal can be passed through the notch filter to obtain the relavent
classification types of EEG signal in given fig 5.3.
We find it pertinent that it is also possible to automatically identify IC1 as containing
an artifactual component by arbitrarily setting a thresholding value anywhere between the
amplitude of features of IC1 and other ICs. However, thisvariant approach is often rigid and
unsuitable to be applied for sporadic and non-stationary signal such as the case of EEG. The
arbitrarily defined threshold may fail to detect the artifactual component or inadvertently
introduced false detection near the boundary of the thresholding value.
We believe that this is a more natural means of emotions recognition, in that the
influence of emotion on facial expression or speech can be suppressed relatively easily, and
emotional status is inherently reflected in the activity of nervous system. The traditional tools
for the investigation of human emotional status are based on the recording and statistical
analysis of physiological signals from the both central and autonomic nervous systems.

Sasi Institute of Technology and Engineering Page 43


A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

Figure 5.3: Filtered ECG signal


That cleaned EEG signal is added with noise or any artifact like ECG, eye blinks,
EKG or EMG etc.. that can be shown in figure 5.4 after classifies the EEG signal with the
sampling frequency and typical amplitude values is shown in figure 5.5.

Figure 5.4: Noise added EEG signal


Sasi Institute of Technology and Engineering Page 44
A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

Figure 5.5: Classification of noise added EEG signal


That noise can be efficiently eliminated by using DTCWT is shown in figure 5.6.

Figure 5.6: Noise eliminated signal

Sasi Institute of Technology and Engineering Page 45


A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

CONCLUSION AND FUTURE WORK


We test a new hybrid procedure for automatic identification and removal of EEG
artifacts substantially improves identification of artifactual components and is found more
reliable than standard thresholding method. Moreover, it promises to be generalizable for
diverse kinds of artifacts, upon selecting proper features and training data. Our system
functions automatically to isolate a distinctly cleaned EEG signal directly from a raw EEG
recording, thus potentially lending itself for applications such as clinical diagnosis or BCI.
We find that the system delivers satisfactory artifact removal without much degrading the
time and frequency resolution of the EEG signal. Future work should serve to identify
features best describe artifactual components extending beyond the present focus on those
arising from the eye blink.

REFERENCES

Sasi Institute of Technology and Engineering Page 46


A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

REFERENCES

[1] E. Tamil, “Electroencephalogram (EEG) Brain Wave Feature Extraction Using Short
Time Fourier Transform”, Faculty of Computer Science and Information Technology,
University of Malaya, 2007.
[2] J. Lee, D. Tan, “Using a Low-Cost Electroencephalograph for Task Classification in HCI
Research”, UIST’06, Monteux, Switzerland, October 15–18, 2006.
[3] G. Molina, “Joint Time-Frequency-Space Classification of EEG in a Brain-Computer
Interface Application”, EURASIPJournal on Applied Signal Processing, Vol. 7, pp. 713–729,
2003.
[4] A. Akrami, “EEG-Based Mental Task Classification: Linear and Nonlinear classification
of Movement Imagery”, in proceedings of the IEEE Engineering in Medicine and Biology
27thAnnualConference Shanghai, China, September 1-4, 2005.
[5] H. BehnamA, A. SheikhaniB, M. Mohammad, M. NoroozianD, P.Golabie, “Analyses of
EEG background activity in Autism disorder with fast Fourier transform and short time
Fourier transform”, International Conference on Intelligent and Advanced Systems,2007.
[6] AbdulhamitSubasi, M. Ismail Gursoy, “EEG signal classification using PCA, ICA, LDA
and support vector machines”, Expert Systems with Applications, Vol.37, pp. 8659–8666,
2010.
[7] Cao, L. J., Chua, K. S., Chong, W. K., Lee, H. P., &Gu, Q. M., “A comparison of PCA,
KPCA and ICA for dimensionality reduction in support vector machine”, Neuron computing,
55, pp. 321–336, 2003.
[8] Subasi, A., “EEG signal classification using wavelet feature extraction and a mixture of
expert model”, Expert Systems with Applications, 32, pp. 1084–1093, 2007.
[9] Ubeyli, E. D., “Analysis of EEG signals by combining eigenvector methods And
multiclass support vector machines”, Computers in Biology and Medicine, 38, pp. 14–22,
2008.
[10] Wang, X., Paliwal, K. K., “Feature extraction and dimensionality reduction algorithms
and their applications in vowel recognition”, Pattern Recognition, 36, pp. 2429–2439, 2003.

Sasi Institute of Technology and Engineering Page 47


A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

Appendix A

PROJECT WORK SCHEDULE

Expec Prerequi Project Associates


Task Durati ted site Topic 15K61A0 16K65A0 15K61A0 15K61A0
No on time Task 467 413 470 472
30th
Oct
2018
3
1 to Literature survey √ √ √ √
weeks
19th
Nov
2018
20th to
2 1 week 28th 1 Study of existing √ √ √ √
Nov techniques
2018
a)
Methodolog √
y Research
29th b) problem

Nov Methodolo formulation
4 2018 gy
3 2,3,4 c) Design of
weeks to implement √
algoritham
3rdJan ation
2019 d) Slot √ √
design
e)implement
√ √
ation
4th Jan
to
4 3 13th 5 Simulatio of designed √ √ √ √
weeks program
Feb
2019
25th
Feb to
5 2 week 12th 4,5 Document √ √ √ √
Mar
2019

Sasi Institute of Technology &Engineering Page A1


A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

APPENDIX B

PROJECT MAPPING WITH PEOS AND POS

Program Educational Objectives Relevance

P2 Design of denoising system from EEG signals Different denoising algorithams.

Program Outcomes Relevance

Engineering knowledge: Apply knowledge of Used knowledge of digital


mathematics, science, engineering fundamentals signal processing for digital
PO1 and an engineering specialization to the solution of image processing for
complex engineering problems. implementing algorithams.

Problem analysis: Identify, formulate, research Performed literature survey on


literature and analyze complex engineering various techniques and
PO2 problems reaching substantiated conclusions using identified the advantages and
first principles of mathematics, natural sciences disadvantages of each of them.
and engineering sciences.

Design/development of solutions: Design A solution was developed for


solutions for complex engineering problems and engineering problem and if
design system components or processes that meet includes the techniques called
PO3
specified needs with appropriate consideration for dual tree complex transform.
public health and safety, cultural, societal and
environmental considerations

Conduct investigations of complex problems: Students analysed and


UseResearch based knowledge and research interpreted data and
PO4
methods including design of experiments, analysis synthesized information while
and interpretation of data and synthesis of doing this project.
information to provide valid conclusions.

Sasi Institute of Technology & Engineering Page B1


A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

Modern tool usage: Create, select and apply Using matlab for simulating
appropriate techniques, resources and modern and implementing the code on
PO5 engineering and IT tools including prediction and EEG signal and to display the
modeling to complex engineering activities with an result.
under- standing of the limitations
The engineer and society: Apply reasoning
informed by contextual knowledge to assess
PO6 societal, health, safety, legal and cultural issues
and the consequent responsibilities relevant to
professional engineering practice.
Environment and sustainability: Understand the
impact of professional engineering solutions in
PO7 societal and environmental contexts and
demonstrate knowledge of and need for sustainable
development
Ethics: Apply ethical principles and commit to While doing project and
PO8 professional ethics and responsibilities and norms documentation followed
of engineering practice ethics.

Individual and team work: Function effectively Individually and as a team


as an individual, and as a member or leader in students worked to implement
PO9
diverse teams and in multidisciplinary settings different techniques in project
to remove noise in eeg signal.

Communication: Communicate effectively on Effectively communicated


complexengineeringactivitieswiththe them to their work through

PO10 engineering community and with society at large, presentations during all
such as being able to comprehend and write reviews understand and
effective reports and design documentation, make answered all quiries.
effective presentations and give and receive clear

Sasi Institute of Technology & Engineering Page B2


A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

Project management and finance: Demonstrate Project managementand


knowledge and understanding of engineering and finance used to implement

PO11 management principles and apply these to one’s project as a member of team
own work, as a member and leader in a team, to to manage project
manage projects and in multidisciplinary
environments

Life-long learning: Recognize the need for and


have the preparation and ability to engage in
PO12
independent and life- long learning in the broadest
context of technological change

Practice in Embedded Systems: An ability to


P0SO1 recognize and adapt to emerging trends in
embedded systems and its applications

Signal & Image Analysis : An ability to perform Signal processing is used to


PSO2 Signal & Image processing in the field of implement this project.
communication

Digital System design: An ability to design a


PSO3
system or process to meet desired needs in VLSI

Sasi Institute of Technology & Engineering Page B3


A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

Project Course Outcomes:

Project Course
Relevance
Outcomes
Able to build
coordination among
project supervisor and Students efficiently coordinated with faculty and did the
CO1
respective students in problem identification..
problem formulation
and idea preparation
Able to survey on
existing and previous
literature on the Literature survey is done by the student by discussing and
CO2
proposed project idea reviewing it by the faculty.
and propose
preferable title.
Able to develop
designated
Methodology is implemented by trails and efficiently done the
CO3 methodology and
work .
design procedure for
intended solution
Able to identify the
challenges faced in
providing intended So many techniques have been executed in order to solve the
CO4
solution and apply problem.
necessary
modifications
Able to enhance team
work ability,
presentation and skills By coordinating team work and by assigning the individual
CO5
for the live work.
demonstration of
proposed project idea
Able to obtain the
results for the
proposed idea, collect Implemented algoritham along with presentation and
CO6
the documented documentation.
evidence and record
the data

(Note: CO1 to 5 are Project work course outcomes of the department)

Sasi Institute of Technology & Engineering Page B4


A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

APPENDIX C
MATLAB SOFT WARE VERSION 2013(a)
The tool used in this project is MATLAB R2011a. MATLAB is a commercial
"Matrix Laboratory" package which operates as an interactive programming environment. It
is a mainstay of the Mathematics Department software line up and is also available for PC's
and Macintoshes and may be found on the Circa vexes. Matlab is well adapted to numerical
experiments since the underlying algorithms for Matlab's built-in functions and supplied m-
files are based on the standard libraries LINPAC and Eispack.

Matlab program and script files always have filenames ending with ".m"; the
programming language is exceptionally straightforward since almost every data object is
assumed to be an array. Graphical output is available to supplement numerical results. Matlab
is a data analysis and visualization tool that has been designed with powerful support for
matrices and matrix operations. As well Matlab has excellent graphics capabilities and its
own powerful programming language. One of the reasons that Matlab has become such an
important tool is through the use of sets of Matlab programs designed to support a particular
task. These sets of programs are called Toolboxes and the particular toolbox of interest. For
example, a matrix, a string, a graph or a figure. Examples of such functions are sin, cos,
imread, imclose etc.

There are many functions in Matlab and are very easy to write our own. A command
is a particular use of a function. We can combine functions and commands or put multiple
commands on a single input line.

Matlab’s standard data type is the ‘matrix’, all data are considered to be matrices of
some sort. Images, of course, are matrices whose elements are the gray values of its pixels.
Single values are considered by Matlab to be 1 x 1 matrices, while a string is merely a 1 x
n matrix of characters, ‘n’ being the string’s length.

Besidesfunctions like imread, figure, plot, input, and output which are used in Matlab
R2011ato read write and show figures or images, some specific function which were used in

Sasi Institute of Technology and Engineering Page C 1


A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

Implementations of this algorithm are listed below:

(1) imread:

A = imread(filename, fmt) reads a grayscale or colour image from the file


specified by the string filename. If the file is not in the current folder, or in a folder on the
MATLAB path, specify the full pathname.

(2) Figure:

figure creates figure graphics objects. figure objects are the individual
windows on the screen in which MATLAB displays graphical output. Figure creates a new
figure object using default property values. To create a figure window that is one quarter the
size of your screen and is positioned in the upper-left corner, use the root object's Screen
Size property to determine the size. Screen Size is a four-element vector: [left, bottom,
width, height].

(3) Plot:

Plot(x, y) creates a 2-D line plot of the data in Y versus the corresponding values in X. If X
and Y are both vectors, then they must have equal length. The plot function plots Y versus X.
If X and Y are both matrices, then they must have equal size. The plot function plots
columns of Y versus columns of X.

(4) Input:

The response to the input prompt can be any MATLAB expression, which is evaluated using
the variables in the current workspace. User entry = input('prompt') displays prompt as a
prompt on the screen, waits for input from the keyboard, and returns the value entered in
user entry. User entry = input ('prompt',’s’) returns the entered string as a text variable rather
than as a variable name or numerical value.

(5) Output:

An output function is a function that an optimization function calls at each iteration of its
algorithm. Typically, you use an output function to generate graphical output, record the

Sasi Institute of Technology and Engineering Page C 2


A Project Report on Dual Tree Complex Wavelet Transform Based EEG Denoising System

history of the data the algorithm generates, or halt the algorithm based on the data at the
current iteration. You can create an output function as a function file, a local function, or
a nested function.

Sasi Institute of Technology and Engineering Page C 3