UNIT I INTRODUCTION TO BCI

Fundamentals of BCI – Structure of BCI system – Classification of BCI – Invasive, Non-

invasive and Partially invasive BCI – EEG signal acquisition - Signal Preprocessing –
Artifacts removal.
Introduction:
Brain-Computer Interface (BCI): A device that utilizes brain activity for direct control of
physical or virtual objects without using muscular activity or body movements.

Applications of BCI:
✦ Improved communication and control for paralyzed and locked-in patients (e.g. stroke,
ALS, spinal injury patients)
✦ Applications in health and safety
E.g. Early detection, diagnosis, and treatment of symptoms
E.g. Alertness monitoring in critical occupations (e.g. night drivers, pilots, railway “engineers”)
✦ Computer-aided education and learning
E.g. Brain-activity based presentation of material?
✦ Augmented cognition (brain-body actuated control)
E.g. Air Force research using hybrid brain-body interfaces for speeding up responses during
flight
✦ Entertainment and Security
E.g. Video games, TV/web browsing for patients,
E.g. Better lie detection devices and “brain fingerprinting.

I. Fundamentals of BCI:

 In principle, any type of brain signal could be used to control a BCI system. The
most commonly studied signals are electrical signals from brain activity measured
from electrodes on the scalp, on the cortical surface, or in the cortex
 A brain-machine interface (BMI) is a device that translates neuronal information
into commands capable of controlling external software or hardware such as a
computer or robotic arm.
 A brain computer interface (BCI), also referred to as a brain machine interface
(BMI), is a hardware and software communications system.
 It enables humans to interact with their surroundings, without the involvement of
peripheral nerves and muscles, by using control signals generated from Electro Encephala
Graphic (EEG) activity.

 It is collaboration between a brain and a device. It enables signal from the brain to

1
 direct some external activity, such as control of a cursor or a prosthetic limb.
 The interface enables a direct communication pathway between the brain and the object
to be controlled.
 In the case of cursor control, for example, the signal is transmitted directly from the brain
to the mechanism directing the cursor, rather than taking the normal route through the
body’s neuromuscular system from the brain to the finger on a mouse.
 By reading signals from an array of neurons and using computer chips and programs to
translate the signals into action.
 BCI can enable a person suffering from paralysis to write a book or control a Motorized
wheel chair or prosthetic limb through thought alone.
 To control a BCI, the user should produce various brain activity patterns which are
captured in form of Electroencephalogram (EEG) and converted to commands by
identifying the patterns by the system.

Figure: General BCI system framework

●Tom, who is paralyzed but is cognitively intact. visits a BCI lab, and sits in front of the
computer screen in his wheelchair, while the lab assistant places an EEG cap (l) on his head. On
the computer screen, the letters and numbers are displayed in matrix form (2).

●Before he can type a letter to his daughter, he needs to train the system to properly record and
respond to his brain signals (3). Rows and columns of the alphanumeric matrix flash randomly,
and Tom is mentally counting how many times the character he wants to type flashes.

●The BCI system then processes his EEG signals, which include distinct brain patterns elicited
by mental counting with respect to the desired characters, to extract the important brain features
by filtering out irrelevant signals. From the extracted brain signal, the BCI system learns how to

2
distinguish what letters he intends to select, so that future attempts to type will be more readily
discernable (4).

●After a few letters, the training is complete and he is able to start typing his message. Again, he
focuses on the letters that he wants to type, and the BCl system applies the classifier that was
built through the training exercise to detect Tom's intention (5). One by one, the letters appear on
the computer screen: DEAR SAMANTHA (6). He is well on his way to writing his letter, using
only his brainwaves.

The purpose of a BCI is to detect and quantify features of brain signals that indicate the user’s
intentions and to translate these features in real time into device commands that accomplish the
user's intent. To achieve this, a BCI system consists of 4 sequential components

 Signal acquisition,
 Feature extraction,
 Feature translation,
 Device output.

These 4 components are controlled by an operating protocol that defines the onset and timing of
operation, the details of signal processing, the nature of the device commands, and the oversight
of performance. An effective operating protocol allows a BCI system to be flexible and to serve
the specific needs of each user.

II. Structure of BCI system:

●The interface comprises EEG acquisition system, feature extraction through data processing
software and pattern classification, and a system capable of transferring the command to external
devices, providing feedback to operator.

●A BCI system whether it is invasive or non-invasive is composed of following phases i.e.

Signal acquisition, signal pre-processing, feature extraction, signal classification and computer
interaction.

Figure: Schematic diagram of EEG based BCI

3
Signal acquisition:
● The ability of a signal acquisition system to measure different phenomena depends
on the transducers to convert the physical phenomena into signals measureable by
the signal acquisition hardware.
● It is the process of digitizing data from the world around us so it can be displayed,
analyzed and stored in a computer.
● Signal acquisition is the measurement of brain signals using a particular sensor
modality (eg, scalp or intracranial electrodes for electro physiologic activity, fMRI
for metabolic activity).
● The signals are amplified to levels suitable for electronic processing (and they may
also be subjected to filtering to remove electrical noise or other undesirable signal
characteristics, such as 60-Hz power line interference).
● The signals are then digitized and transmitted to a computer.
Signal pre-processing:
● Signal pre-processing is also called as signal enhancement.
● After data acquisition we first preprocess the data to extract the feature.
● It usually intensifies the signals and the upgrade signal to noise ratio (SNR).
● The general step in the preprocessing is band pass filtering which were designed to
remove DC bias and high frequency noises.
● Spatial filtering and frequency filtering may be used in the EEG signal generation
for preprocessing.
Feature extraction:
● Feature extraction is the process of analyzing the digital signals to distinguish
pertinent signal characteristics (ie, signal features related to the person's intent) from
extraneous content and representing them in a compact form suitable for translation
into output commands.
● These features should have strong correlations with the user's intent. Because much of
the relevant (ie, most strongly correlated) brain activity is either transient or
oscillatory.
● The most commonly extracted signal features in current BCI systems are time-triggered
EEG or ECoG response amplitudes and latencies, power within specific EEG or

4
 ECoG frequency bands, or firing rates of individual cortical neurons.
▪ Environmental artifacts and physiologic artifacts such as electro myographic signals
are avoided or removed to ensure accurate measurement of the brain signal features
● When the data is too large to be processed, the data will be transformed into a reduced
representation set of features.
● The process of transforming the input data into the set of features is called
feature extraction.
● The goal of feature extraction is selecting suitable data for the subsequent classification
or detection of features needed to design.
● It is the process of collecting discriminative information from a set of samples.

● Transforming the input data into the set of features is called feature
extraction.

Feature Translation:
● The central element in each BCI is the classification module which is also referred
to as translation algorithm.
● It simply converts electrophysiological input from the user into output that
controls external devices.
● The translation algorithm is an important stage in the signal processing module of
the BCI system.
● It is responsible for translating the extracted signal features into device commands
that performs the user’s intent.
● Whatever the nature is, a translation algorithm changes signal features into device
control commands.
● The first part of signal processing simply extracts specific signal features.
● The extracted signal features may be classified on both frequency and shape
features based on linear methods or non-linear methods like the neural networks.
● The resulting signal features are then passed to the feature translation algorithm,
which converts the features into the appropriate commands for the output device
(ie, commands that accomplish the user's intent).

5
 ● For example, a power decrease in a given frequency band could be translated into
an upward displacement of a computer cursor, or a P300 potential could be
translated into selection of the letter that evoked it.
● The translation algorithm should be dynamic to accommodate and adapt to
spontaneous or learned changes in the signal features and to ensure that the user's
possible range of feature values covers the full range of device control.

Device Output:
● The commands from the feature translation algorithm operate the external device,
providing functions such as letter selection, cursor control, robotic arm operation,
and so forth.
● The device operation provides feedback to the user, thus closing the control loop.
● In computer programming, an application programming interface (API) is a set of
subroutine definitions, communication protocols, and tools for building software.
● In general terms, it is a set of clearly defined methods of communication among
various components.

III. Classification of BCI:

Figure :Classification of BCI system

6
Brain Signal Pattern:

Signals rely on different arrangements of impulses to communicate with the specific application.

An example of a brain signal is P300 event-related potentials (ERPs) that reach a maximum
positive peak in voltage about 300 ms after a stimulus onset (e.g., through the so-called oddball
paradigm).

Compared with other brain signals, P300 ERPs require little initial training-a huge advantage as
compared to other brain signal types. P300 can be evoked by visual, auditory, tactile, and even
olfactory or gustatory paradigms.

Another widely used brain signal type is steady-state evoked potentials (SSEPs). which are the
electrical activity of the brain in response to stimulation of specific sensory nerve pathways, as
distinct from spontaneous potentials.

There are three kinds of SSEPs:

▪Steady-state auditory evoked potentials (SSAEPs), usually recorded from the scalp but
originating at the brainstem level.

▪Steady-state visually evoked potentials (SSVEP) found within the occipital lobe of the brain
caused by focusing on a steady pattern of visual stimuli, such as a regularly flashing light;

▪Steady-state somatosensory evoked potentials (SSSEPs), a sinusoidal electrophysiological brain

response elicited from mechanical vibrotactile stimulation delivered to the glabrous skin (e.g.
fingertip). which is modulated by selective spatial attention.

Among many SSEPs, SSVEP-based BCIs that utilize flickering light sources with different
frequencies from either LED or LCD displays are well studied, because they could provide the
quickest and most reliable path to communication.

For example, when the user concentrates on a target stimulus with a certain frequency, that same
frequency is synchronized to a certain area according to the modality (e.g., visual cortex for
visual stimuli), and the amplitude at the frequency of the target stimulus is higher than those of
the non target stimuli.

The BCI system uses these distinct brain patterns to perform different actions (e.g. directions,
on/off, and typing) according to the user's selective attention.

▪A lesser used brain signal method is slow cortical potentials (SCPs), which are shifts in the
cortical electrical activity lasting from several hundred milliseconds to several seconds.

7
▪Finally, there is sensorimotor rhythm (SMR), involving event-related
synchronization/synchronization.

▪(ERD/ERS), brain activity associated with imagining motor behavior.

IV. Signal Acquisition or Recording Methods:

Types of Recording Methods

1. Invasive

2. Partially Invasive

3. Non-Invasive

Figure: Types of Recording BCI

BCIs require a neuroimaging or neurophysiological device to acquire and transmit the brain
signals from brain to computer. In general, neuroimaging methods are categorized by
invasiveness the recording methods, but can be further classified by spatial/temporal resolution,
direct indirect measurement, and complexity/price. Each recording technique has strengths,
weaknesses, and specific uses that help researchers decide which device is relevant to their
study: Figure visually compares different recording methods discussed in more detail in the
following section.

1. Noninvasive Recording Methods:

Safest and low cost type

The detection of signals is done by some electrodes placed on the scalp.

These devices have weaker human brain signals than other BCI devices due to the skull but
placing the electrodes is easy as well as portable.

8
A non-invasive recording technique uses sensors placed on the skin, such as the scalp or
machinery that surrounds the cranium in whole.

Two types of non-invasive recording methods,

● Direct measures that detect electrical (e.g., EEG) or magnetic activity (e.g. MEG) of

the brain.

● Indirect measures of brain function reflecting brain metabolism or hemodynamics of the brain
(e.g., fMRI, fNIRS, and PET) that do not directly characterize the neuronal activity. Unlike
invasive recording methods, these non-invasive techniques do not require surgery, internal
chemical or machine implantation, or needle insertion in order to receive and record neural
activity.

The safest and most prevalent of these methods is non-invasive.

This method requires equipment that either touches the scalp such as electroencephalogram
(EEG) or near-infrared spectroscopy (fNIRS) or is otherwise outside of the cranium such as
functional magnetic resonance imaging (fMRI), magnetoencephalography (MEG), or positron
emission tomography (PET).

▪Electroencephalography EEG

▪ 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.

▪Magnetoencephalography (MEG)

▪ MEG detects the tiny magnetic fields created as individual neurons "fire" within the brain.

▪ It can pinpoint the active region with a millimeter, and can follow the movement of brain
activity as it travels from region to region within the brain.

•Functional Magnetic Resonance Imaging (fMRI)

• It exploits the changes in the magnetic properties of haemoglobin as it carries oxygen.

• Activation of a part of the brain increases oxygen levels thereby increasing the ratio of
oxyhaemoglobin to deoxyhaemoglobin.

9
2. Invasive Methods:

● Special devices called invasive BCI devices are used to capture signals from the brain.
These devices are inserted into the human brain by a critical surgery.
● Devices that are capable of detecting signals from single area is called single unit while
from multiple areas of brain are called multiple units.
● These devices has the quality of forming high quality signals but also possess the risk of
forming scars.
● Unlike non-invasive methods, invasive techniques require surgery in order to place
signal receptors within the brain. Examples of invasive recording methods are
intracortical neuronal recording and electrocochleography (ECoG).
● Invasive BCls are implanted directly into the grey matter of the brain by neurosurgery.

3. Partially Invasive:

● Devices are inserted in the skull on the top of human brain.

● These devices have bit weaker quality of human brain signals than invasive BCls and
have less risk of forming scar tissue.
● It is another brain signal reading process which is applied to the inside the skull but
outside the grey matter.
● Electrocorticography (ECoD is the example of partially invasive BCI.
● An electrocorticograph (ECoG) records the activity of the brain inside the skull, but
from the surface of the membranes that protect it.

V. Signal Pre-processing methods in BCI:

To extract features, it is necessary to pre-process first the data. Three steps are necessary to
achieve this goal:

 Referencing,
 Temporal filtering and
 signal enhancement.

1. Referencing:

• In the case of EEG recordings from the cortex or from the scalp, these recordings are obtained
using, in general, different electrodes on different positions.

• Since the brain activity voltage measured by a given electrode is a relative measure, the
measurement may be compared to another reference brain voltage situated on another site.

10
• This results in a combination of brain activity at the given electrode, brain activity at the
reference site and noise. Because of this, the reference site should be chosen such that the brain
activity at that site is almost zero.

• Typically, the nose, mastoids and earlobes are used. In general, there are three referencing
methods

i. Common reference:

•The common reference technique is widely used in BCIs.

•This method uses one common reference for all electrodes.

•In general, the site of this reference is situated at large distance from all electrodes. The activity
at the reference site influences all measurements equally, and differences between electrode
measurements still contain all information needed.

ii. Average reference:

• The average reference subtracts the average of the activity at all electrodes from the
measurements.

• This method is based on the principle that the activity at the whole head at every moment sums
up to zero.

• Therefore, the average of all activity represents an estimate of the activity at the reference site.

• Subtracting this average produces in principle a dereferenced solution.

• However, the relatively low density of the electrodes and the fact that the lower part of the head
is not taken into account, bring some practical problems along.

iii. Current source density (CSD):

• The current source density (CSD) is used in many BCIs. It is "the rate of change of current
flowing into and through the scalp".

• This quantity can be derived from EEG data, and it may be interpreted as the potential
difference between an electrode and a weighted average of their surrounding electrodes.

• The CSD can be estimated by computing the laplacian. The laplacian computes the sum
of the differences between an electrode and its neighbours.

• A problem with this estimation is that it is actually only valid when the electrodes are in a two
dimensional plane and equally distant.

11
2.Temporal filtering in BCIs:

• The brain signals are naturally contaminated by many internal and external noises.

• They can be removed using simple filters.

• The relevant information in BCIs is found in the frequencies below 30Hz.

• Therefore, all noise with higher frequencies (e.g. noise from the electrical net has a fixed
frequency of 50Hz or 60 Hz) can be removed using FIR low pass filter.

• Specific frequency bands may also be selected using FIR bandpass filters.

VI. Signal enhancement methods in BCI design:

•The choice of a suitable enhancement technique is dependent on several factors such as the
recording technology, number of electrodes, and neuro mechanism of the BCI.

•Among seventeen pre-processing methods given by, we describe here briefly only six methods
which are the most applied in BCI designs:

1.Spatial filters - Referencing methods:

The proper selection of a spatial filter for any BCI is determined by the location and extent of the
selected brain control signal and of the various sources of EEG or non- EEG noise.

2.Common average referencing (CAR):

• Common-average or "reference-free" recording has been suggested as a solution to the problem

of the reference electrode. average referencing involves recording in bipolar fashion from a
number of electrodes, all referred to a single site.

• One then calculates the grand mean EEG waveform, by averaging across electrodes, and
subtracts the result pointwise from the EEG recorded at each electrode.

• Activity recorded by the reference electrode is theoretically of equal magnitude in the mean and
individual electrode waveforms.

• Consequently, the effect of the reference electrode should be eliminated from each recording
electrode's output when the common-average waveform is subtracted.

3.Surface Laplacian (SL):

• The SL is defined as the 2nd order spatial derivative of the surface potential.

12
• Due to its intrinsic spatial high-pass filtering characteristics, the SL can reduce the volume
conduction effect by enhancing the high-frequency spatial components, therefore can achieve
higher spatial resolution than surface potentials.

4. Principal component analysis (PCA):

• The PCA is a linear mapping 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.

• Depending on the field of application, it is also named the discrete Karhunen-Loève transform
(KLT), the Hotelling transform or proper orthogonal decomposition (POD).

• The PCA reveals the internal structure of the data in a way which best explains the variance in
the data.

• If a multivariate dataset is visualised as a set of coordinates in a high-dimensional data space (1

axis per variable), ICA supplies the user with a lower-dimensional representation.

5.Independent component analysis (ICA):

• The more important artifacts in BCIs are generated by muscles and eyes blink.

• Classical automatic methods for removing such artifacts can be classified into rejection
methods and subtraction methods.

6. Rejection methods:

It consist of discarding contaminated EEG, based on either automatic or visual detection can be
used in the BCI applications framework. Their success crucially depends on the quality of the
detection.

7. Subtraction methods:

It based on the assumption that the contaminated EEG is a linear combination of an original EEG
and other independent artefact signals generated by the muscles and eyes blink.

• The original EEG is hence recovered by either subtracting separately recorded artefact-related
signals from the measured EEG, using appropriate weights or by applying recent approaches
for artefacts rejection: such as independent component analysis (ICA), peak elimination and
fixed bandpass FIR filter based approach is the more used technique.

• It is a computational method for separating a multivariate signal into additive subcomponents

supposing the mutual statistical independence of the non- Gaussian source signals.

13
• It is a special case of blind source separation (BSS). ICA is particularly efficient when the EEG
and the artefacts have comparable amplitudes.

• For more details about their advantages, their limitations and their applications for the removal
of eyes activity artefacts, refer to.

8. Common spatial patterns (CSP):

The CSP is a technique used to find the common projection matrix that decomposes the different
classes of single trial EEG datasets, and more specifically to find spatial structures of event-
related (de)synchronization (ERD/ERS) in a EEG context. Such matrix maximizes the
differences between the classes.

VII. Artifacts Removal in EEG:

 Artifacts in EEG recordings can be caused by eye blinks, eye movement, muscle and
cardiac noise, as well as nonbiological sources (e.g., power-line noise).
 A problem arises if the artifacts generated by the subject are used to control the BCI
system, because this violates the definition of a BCI as a nonmuscular communication
channel.
 Furthermore, subjects with degenerative diseases would eventually lose this ability.
 For instance, these artifacts could be a voluntary or involuntary blinks or muscle
contractions when the task is presented.
 Additionally, involuntary muscle or ocular activity might obscure the actual EEG
 signal, obstructing measurement of the features used to control the system.
 Electromyographic (EMG) activity tends to overlap EEG from 8 Hz upwards, whereas
electrooculographic (EOG) activity overlaps in the range 0–12 Hz (e.g., the μ-rhythm has
a frequency of 8–12 Hz).

1.Regression Methods:

 The traditional method for removing artifacts from EEG is the regression methods.
 It is applied under the assumption that each channel is the cumulative sum of pure EEG
data and a proportion of artifact.
 Regression analysis first defines the amplitude relation between reference channel
 and EEG channel by transmission factors, and then subtracting the estimated artifacts
from EEG.
 Thus, this algorithm requires exogenous reference channels (i.e., EOG, ECG) to omit
different artifacts. When dealing with ocular activity, the EEG data can be obtained as:

EEGcor = EEGraw − γF(HEOG) − δF(VEOG) (1)

14
where γ and δ depend on the transmission coefficient between EOG and EEG, and EEGcor and
EEGraw represent corrected EEG data and raw EEG data, respectively. HEOG and VEOG
denote the recordings from horizontal and vertical EOG channels.

2.Wavelet Transform:

 Wavelet transform, transforming a time domain signal into time and frequency domain,
has good time-frequency features relative to Fourier transform due to the better tunable
time-frequency tradeoff and superiority of non-stationary signal analysis.
 The transformation is accomplished by selecting the subsets of the scales ‘j’ and the time
shift ‘k’ of the mother wavelet ψ(t). Mathematically:
Ψj,k (t)= 2j/2 ψ( 2j t-k) …. 1

Where j and k are integers. Then the wavelets transform can be performed by

Wψ =< f, ψj,k > ….2

Which means the inner product of time-domain signal and wavelet function.

3. BSS:

 The BSS method includes a variety of unsupervised learning algorithms without prior
information and extra reference channels.
 The general methodology of BSS can be described as follows.
 Let X be observed signals obtained from scalp electrodes.
 Also, let S be the source signals which includes original signals and artifacts. These
source signals are linear mixed by an unknown matrix A:

X=AS,

To get the observed signals. The BSS algorithm is a reversed version.

U=WX,

where U is the estimation of sources and the W is the reverse mixing of X.

 Then components representing the artifacts are removed and the remaining components
reconstruct EEG data to achieve the purpose of denoising.

15