INTEGRATION OF SIGNAL PROCESSING METHODS
INTO EEG/ERP SYSTEM
Petr Ježek and Roman Mouček
Department of Computer Science and Engineering, University of West Bohemia
Univerzitní 8, 306 14 Pilsen, Czech Republic
Keywords: EEG/ERP signal processing, EEG/ERP experiment, Independent component analysis, Principal component
analysis, FastICA, Matching pursuit, Plug-in, PluginEngine, JERPA, JUIGLE.
Abstract: This paper deals with the processing of signals from EEG/ERP experiments. Selected methods for
EEG/ERP signal processing are shortly presented. Since these methods are usually implemented for various
platforms and within various tools, a new designed and developed software system, which allows users
running them uniformly, is presented. The methods are added and processed as plug-ins. It ensures a high
modularity and flexibility of the system. The main system components (JERPA, JUIGLE) are introduced
and shortly described.
1 INTRODUCTION
In our department (Department of Computer Science
and Engineering, University of West Bohemia,
Pilsen) we specialize in the research focused on
attention, especially attention of drivers or seriously
injured people. We use the methods of
electroencephalography (EEG) and Event-Related
potentials (ERP). We collaborate with many
institutions e. g.: (Czech Technical University in
Prague, University Hospital in Pilsen, Škoda Auto,
Inc…).
Our research group performs experiments
according to designed scenarios in a well equipped
EEG/ERP laboratory. Data/metadata from
experiments are stored into developed EEG/ERP
portal (Ježek and Mouček, 2010). When
data/metadata from experiments are obtained we
process them using various signal processing
methods. The existing methods are implemented by
various vendors. These methods are usually platform
dependent and use various user interfaces; they
usually need to install additional frameworks or
virtual machines for their running.
Since the work with these tools is unfriendly we
have decided to implement own framework. This
framework integrates implemented methods – they
are added as plug-ins. Integration with the developed
portal will be ensured.
2 METHODS FOR EEG/ERP
SIGNAL PROCESSING
EEG/ERP signal obtained from a recording device is
often distorted by artifacts originated from muscle
activity, eyes movement, breathing, etc. Some
methods used for signal separation, signal
decomposition and removal of artifacts are described
in this section.
2.1 Independent Component Analysis
The goal of independent component analysis (ICA)
(Hyvärinen, Karhunen and Oja, 2001) is to find
a linear representation of nongaussian data so that
the components are statistically independent, or as
independent as possible. Such a representation
seems to capture the essential structure of the data in
many applications, including feature extraction and
signal separation.
2.1.1 FastICA
FastICA (Oja and Zhijian, 2006) is algorithm that
solves some difficulties of ICA. The most important
is very fast convergence, there is no need to set a
step size. The algorithm finds independent
components directly; they are calculated one after
another.
563
Ježek P. and Mou
ˇ
cek R..
INTEGRATION OF SIGNAL PROCESSING METHODS INTO EEG/ERP SYSTEM .
DOI: 10.5220/0003165305630566
In Proceedings of the International Conference on Health Informatics (HEALTHINF-2011), pages 563-566
ISBN: 978-989-8425-34-8
Copyright
c
 2011 SCITEPRESS (Science and Technology Publications, Lda.)

2.2 Principal Component Analysis
Principal component analysis (PCA) (Jolliffe, 2002)
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.
2.3 Matching Pursuit
Algorithm (Mallat and Zhang, 1993) decomposes
any signal into a linear expansion of waveforms that
are selected from a redundant dictionary of
functions. These waveforms are chosen in order to
best match the signal structure. The dictionary is
often based on Gabor functions.
3 SYSTEM SPECIFICATION
3.1 System Context
Because of uncomfortable work with existing
methods for EEG/ERP signal processing we have
decided to design and implement a custom software
tool (called the system in the following text). The
system should serve not only for our purposes but
also for our collaborators or users interested in
signal processing.
We believe that the developed system will help
to remove difficulties with using signal processing
methods and will increase efficiency of
neuroscientific research.
3.2 System Requirements and Users
The system is primarily designed for our department
and our collaborators, but we suppose that the other
users interested in signal processing will welcome
this tool.
Hence the system has to be suitable for users
who have only basic computer knowledge. These
users want to work with a simple interface with
possibilities to install updates and new plug-ins
(methods for signal processing). In addition,
advanced users can implement their own signal
processing methods and provide them within the
system as plug-ins.
3.3 Project Scope
The system is developed as a standalone desktop
product with data access into various data sources
(e.g.: developed EEG/ERP portal).
The system and contained plug-ins are based on
the Java platform and they are provided as an open
source. It ensures a higher availability for potential
users and faster development of new plug-ins. The
main parts of the system are:
 GUI framework – support for creating graphical
components presenting output of implemented
methods.
 Plug-in engine – engine for installing new
methods as plug-ins into the system
 Error report engine – framework for sending
errors occurred in the system (e.g. email, log files, or
SMS)
 Data store engine – connection of the system to
existing EEG/ERP portal (or another databases).
 Localization support – user interface is translated
into a user language
3.4 User Roles
Since the system processes personal data of tested
subjects it is necessary to ensure the system security.
The system of user roles is developed. User has
to create his/her user account, system administrator
assigns user roles.
In the developed prototype we have one role
“researcher”. Researcher has administrator
privileges, but the system is prepared to add new
roles. With new roles we can restrict privileges; the
user is prohibited from installing new plug-ins or
processing non public data.
4 SYSTEM DESIGN
AND REALIZATION
4.1 System Architecture
The system is a layered architecture. This
architectural style is directly supported by used
technologies and ensures high level of abstraction as
well as a long term existence of the system as open
source.
The next subsections describe the main modules
of the system.
HEALTHINF 2011 - International Conference on Health Informatics
564

4.1.1 JERPA
JERPA (Java Event-Related Potential Analysis) is
the main component of the system encapsulating
other components (Figure 1).
Figure 1: JERPA Component Model.
Data and application layers are based on jERP
studio (Řondík, 2008). The data layer allows users to
read raw EEG/ERP data and corresponding metadata
in real time. The application layer provides an
interface for creating EEG/ERP signal processing
methods and corresponding plug-ins. It also runs a
graphical user interface.
The presentation layer of the JERPA component
is based on the JUIGLE library (see Section 4.1.2).
It includes the following perspectives:
 Signal processing (EEG/ERP signal visualization
and recording, epochs and artifacts marking,
baseline correction, epochs averaging, results
exports, and import from other projects)
 Algorithm Manager (list of available methods,
classification and sorting of methods, selection of
methods and their application to EEG/ERP signal)
 JERPA settings (language selection,
activation/deactivation of plug-ins, etc.)
4.1.2 JUIGLE
JUIGLE (Java User Interface Graphic Library
Environment) is a library initially providing the
system user interface. It is based on the three layer
architecture and allows drawing charts, which
contain the results of implemented methods.
JFreeChart library is used; JUIGLE provides an
interface for a communication with JFreeChart.
Then a user can put results into a selected chart by
implementing the interface.
JUIGLE is also designed to simplify the system
development. The library features include e.g.
reduction of writing the same source code for GUI, a
plug-in system, a database engine, an engine for
reporting errors and exceptions, or a framework
allowing component communication using the
observer pattern. Figure 2 presents the JUIGLE
component model.
JUIGLE contains a database connector for
loading input data. The connector is an abstract
interface, so data can be loaded from a relational
database, XML or another data source. Then an
implementation of the specific connector is
necessary. JUIGLE also supports the project
management. A user can save unfinished projects
and continue in the work later.
Figure 2: JUIGLE Component Model.
4.1.3 PluginEngine
Because we need the system to add new methods
easily, we have also developed a plug-in engine.
Plug-ins are java libraries which implement interface
provided by PluginEngine. PluginEngine loads plug-
ins dynamically according to XML configuration
file. PluginEngine can actualize plug-ins, set
properties (hidden, forbidden, active), install new
plug-ins or sort them.
Plug-in developers define method input and apart
from common table output they can also provide
output in a chart. A developer only selects a chart
type, defines axis values, and the framework ensures
the rendering of selected chart.
INTEGRATION OF SIGNAL PROCESSING METHODS INTO EEG/ERP SYSTEM
565

Figure 3: Component model of the system.
The methods shortly presented in Section 2 were
implemented as the plug-ins.
4.1.4 Additional Modules
The system contains other modules providing
additional functionality (e.g. JERPALang for JERPA
localization, JERPATest for methods testing, Java
Mail for email communication, NimROD for setting
Look & Feel, or Log4j for logging information and
error outputs. Figure 3 shows a simplified
component model of the system.
5 CONCLUSIONS
In this paper we presented difficulties with the
processing of data from EEG/ERP experiments.
Because implementations of existing methods are
often not suitable for their usage in real world
applications (they are difficult to control,
implemented for inconsistent target platforms, or
have insufficient documentation) we designed and
particularly implemented own system.
The presented system combines research in
EEG/ERP with modern software engineering
approaches. It increases research efficiency and
makes its results widely available.
The system provides the implementation of
introduced methods. Users can add their methods for
EEG/ERP processing as plug-ins. Since the system
is developed using common open-source
technologies, an extension of the system using plug-
ins is easy.
As the next step we are preparing an integration
of the presented system with already developed
EEG/ERP portal (Ježek and Mouček, 2010). We are
going to ensure a complex way of EEG/ERP signal
processing, which includes managing EEG/ERP
experiments and data/metadata from them,
EEG/ERP signal processing and distribution of
results to collaborators or interested researchers.
ACKNOWLEDGEMENTS
The work was supported by the Ministry of
Education, Youth and Sport of the Czech Republic
under the grant ME 949.
REFERENCES
Ježek, P., Mouček, R., 2010. Database of EEG/ERP
experiments.., 2010. In Healthinf 2010 – Proceeding
Third International Conference on Health Informatics.
Valencia Spain.
Hyvärinen, A., Karhunen, J., Oja, E., 2001. Independent
Component Analysis. John Wiley & Sons. Hardcover.
Jolliffe, I. T. 2002. Principal Component Analysis,
Springer.
Oja, E., Zhijian Y., 2006. The FastICA Algorithm
Revisited: Convergence Analysis, IEEE transactions
on neural networks.
Mallat, S. G., Zhang, Z. 1993. Matching Pursuits with
Time-Frequency Dictionaries, IEEE Transactions on
Signal Processing.
Řondík, T., 2008. Thesis. Software for ERP processing,
Pilsen.
HEALTHINF 2011 - International Conference on Health Informatics
566