• Nie Znaleziono Wyników

Szczepan PASZKIEL* CONCEPT OF EXPERT SYSTEM INTERPRETING CORRECTNESS OF MEASUREMENT AND METHOD OF THE EEG SIGNAL ANALYSIS FOR NEEDS OF THE BRAIN-COMPUTER INTERFACE

N/A
N/A
Protected

Academic year: 2021

Share "Szczepan PASZKIEL* CONCEPT OF EXPERT SYSTEM INTERPRETING CORRECTNESS OF MEASUREMENT AND METHOD OF THE EEG SIGNAL ANALYSIS FOR NEEDS OF THE BRAIN-COMPUTER INTERFACE"

Copied!
8
0
0

Pełen tekst

(1)

__________________________________________

* Opole University of Technology.

Szczepan PASZKIEL*

CONCEPT OF EXPERT SYSTEM INTERPRETING CORRECTNESS OF MEASUREMENT AND METHOD OF THE EEG SIGNAL ANALYSIS FOR NEEDS OF THE BRAIN-COMPUTER INTERFACE

The concept of construction of the expert system interpreting correctness of measurement and the method of the EEG signal analysis for needs of the brain-computer interface (BCI) are described in the article. The general orientations related with the methodology of creating the expert systems based on the knowledge base (KB) are characterised. Also, the brain-computer interface technology was described, which has been recently gaining more and more popularity within the scope of its application in control processes.

KEYWORDS: EEG signal, expert system, data analysis

1. INTRODUCTION

While designing interfaces on the basis of the brain-computer technology it is crucial to identify and match correctly the neuroinformatic tool for the analysis of the data flowing for a human brain, collected during performing of the electroencephalography (EEG) or functional magnetic resonance (fMRI). An example of the principle assumption of the fMRI measurements is studying of correlations of cognitive, functional and sensory processes. Through conducting the data analysis using statistical tools it is possible to identify these areas of the human brain, which are responsible for reacting to an external stimulus.

Currently, the control mechanisms constructed on the basis of the brain- computer technology [1] are used more and more widely. In order to connect correctly the measuring device with the control object or the computer application, it is essential to select correctly the tool for identification and verification of the measurement data, e.g. the electroencephalography signal. For this purpose the interpreting expert system suggested in this article and focused at the EEG data analysis [2] may be helpful.

(2)

2. EXPERT SYSTEMS

Computers may feature artificial intelligence if they are programmed in the relevant way. Taking into consideration the present evolution within this realm in future we can expect intelligent robots, which will be able to ‘think’

independently, in order to solve many complex problems [3].

The expert system is a computer program that uses the knowledge and inference procedures in order to solve numerous different problems, which are highly complex and require substantial expert knowledge. The crucial factor at elaboration of the expert system is the knowledge that provides the relevant level of the generated expertise. The expert systems are applied chiefly in order to:

lower the costs of realized expertises, lack of experts in many realms, which is currently visible especially in the neuroinformatics, etc. Moreover, the expert systems work much faster than human experts, they are highly: consistent, objective and precise.

3. BRAIN-COMPUTER TECHNOLOGY

The brain-computer interface technology was created on the basis of accomplishments in some fields of science, which among others include:

medicine, computer science as well as automatics and robotics [4]. To understand the essence of operation of the brain-computer interfaces correctly the knowledge concerning human brain development and its structure is indispensable [5]. During the life span of a human being our brain develops dynamically already in the mother’s womb reaching the pleated cortex structure in the 9-th month of the fetal life. For the next consecutive years from the date of birth there is very rapid growth of the network of connections among neurons that are being formed. Synapses are formed, which at a mature human being are determined by the number: 10^(14), which, as it is demonstrated by academic studies, allows for performing of about 10^(15) operations per second. Each neuron in the human brain consists of: dendrites, an axon, soma and nucleus.

From the point of view of designing of the brain-computer interfaces division of neurons into pyramidal cells and interneurons. The pyramidal cells are positioned perpendicularly to the brain cortex surface and they are in large extent the main source of the EEG signal registered by the encephalographs [6].

In a human brain we deal with resting potential (-70 mV), action potential (+30 mV) and equilibrium potential. Due to the way of correlation of neurons being the fundamental source of information in the brain-computer technology, the notion of a neurotransmitter is important. A neuro relay, or rather neurotransmitter (e.g.: GABA, acetylcholine, noradrenaline, dopamine, serotonin) is used for changing an electric signal into a chemical one in a

(3)

synapse and for transferring this signal from one presynaptic cell to the other one postsynaptic.

In the BCI technology two measurement methods are used: invasive and non- invasive. The non-invasive method used more frequently by far, especially in case of commercial applications, is based on the above mentioned electroencephalography. In the academic publications there are also to be found implementations based on fMRI, however, they are more complex and expensive in realization than the standard methods based on the EEG.

On the other hand, the invasive method depends on direct positioning of measurement electrodes on the surface of the brain cortex and due to substantial level of interference into human organism is used only in justified cases (advanced organism dysfunctions). Currently, two firms distributing and designing products based on the BCI technology are in the market of commercial solutions, i.e.: Emotiv Inc. and NeuroSky. The most popular device of the Emotiv company is: Emotiv EPOC+ NeuroHeadset, while of the NeuroSky company: NeuroSky MindWave Mobile.

In order to be able to take the EEG signal measurements the measurement electrodes should be skilfully positioned on the surface of the head. It is realized on the basis of the 10-20 system defined by the International Federation of Clinical Neurophysiology (IFCN). The distances between the particular electrodes are determined by the percentage values determined on the basis of the boundary points. In medical implementations we deal with two types of measurement electrodes: mushroom and disc. The devices made by Emotiv Inc.

work on the basis of the disc electrodes. The electrodes register the EEG signal, which is a record of a brain electric activity variable over the time [7]. A few kinds of wave are separated in the EEG signal: alpha in the range: 8-13 Hz, theta in the range: 4-9 Hz, beta in the range: 13-30 Hz, gamma in the range: 26-100 Hz, delta in the range: to 4 Hz. The alpha waves are most frequently related to vigil, theta with meditation, beta with daily routine, gamma can be noticed while realizing the perception process and remembering; whilst delta are the waves characteristic for deep sleep [8]. Unfortunately the measurement of the EEG signal with the non-invasive method using the equipment based on electroencephalography (EEG) is burdened with artefacts: technical and biological. The technical artefact instead are directly connected with the measurement environment, operating computers, devices, electric mains, etc.

The biological artefacts result instead directly from the features of the particular organism (high blood pressure, increased heart beat, nervous twitch, etc.), which is subject of examining for the needs of control processes [9].

(4)

4. NEUROINFORMATICS TOOLS

Neuroinformatics is a highly interdisciplinary field of science targeted to use analysis methods in neurobiology, modelling and measurements, which originated from physics with use of the IT technologies. The beginning of neuroinformatics is dated to 1979, when the first computer system pre-defined for centralized gathering and management of medical images was created by Prof. Heinz Lamke. It is highly probable that the broad spectrum of applications of IT tools over the longer time perspective will allow for the processed data to permit for modifications of healthy brain models in order to simulate diseases.

This will allow the scientists to study mechanisms causing the brain disorders, which could substantially accelerate medical studies in this respect. There is the organization called the International Neuroinformatics Coordination Facility acting in the world, which coordinates development of neuroinformatics and was founded in 2005 in Sweden. Within the scope of works of the above mentioned organization the DataSpace project is realized concerning providing of the space for exchange of the data and brain operation simulations. SoftwareCenter is the software source for neuroinformatics. NineML is an independent language for equivalence of description of the neuron models.

Within the commonly used tools applied in neuroinformatics we can enumerate among others: Laplace’s filters, thanks to which the mean values of the potential flowing in the EEG signal can be averaged. Methods of analysis:

ICA, PCA, Higher Order Statistic, Secound Order Statistic. The following toolboxes are available in the Matlab environment: EEGLAB, WFDB etc.

The data subject to acquisition by the value measurement on the human head surface can also be used for elaboration of nerve cell models, including population and unitary models. E.g. the Hodgkin’s-Huxley’s model is the one of the synaptic membrane. The FitzHugh’s-Nagumo’s and Hindmarsh’s-Rose’s models are the ones made on the basis of the Hodgkin’s-Huxley’s model. The next models are: Lopes da Silva's model, which presents generating of alpha rhythm oscillations; Jansen’s-Rit’s model is a development of Lopes da Silva’s model, in which a larger population of neurons is analysed; David’s-Friston’s model replays spectral properties of the EEG signal in the wide frequency band.

5. PROPOSED EXPERT SYSTEM

On the basis of the information concerning operation of devices based on the brain-computer technology and accessible neuroinformatic tools a concept of creating an interpreting expert system was proposed.

Within the framework of conducted studies and a number of experiments with practical implementation of the BCI technology, an outline of the expert

(5)

system was elaborated, which includes: the knowledge data base, knowledge gathering module, concluding module, presentation module and user interface, through which an expert, knowledge engineer and user communicate with the system. The knowledge gathering module is a set of methods recorded in a computer program, which enable archiving the knowledge in the form of facts and rules. In this case this is the EEG signal characteristic for the particular states of thoughts. The concluding module is a set of orders and methods incorporated in the program, which render it possible to utilize the rules and data included in the knowledge data base. The module presenting this software objected to interpret the obtained results for the person busy with constructing the brain-computer interfaces. In the proposed expert system the expert system knowledge will comprise facts, these are the EEG signals classified relevantly.

An important observation is the fact that the level of generated expertises is the function g(x) (1), where: skdb – size of knowledge base, qkdb – knowledge quality in the data base.

kdb

kdb

q

s x

g ( )  

(1)

In order to elaborate the expert system as the tool supporting work of the constructors of the devices based on the brain-computer technology a relevant problem analysis is necessary. In the result of realized insight it is possible to state that there is a large demand for this type of tool. Polled potential future users of the system defined the functions and expectations. The principal expected functionality on behalf of the users was the possibility of analysis of the data collected from a device such as an electroencephalograph is and then via utilization of the neuroinformatic tools, the possibility of comparison of the results with patterns of the EEG signal behaviours included in the data base.

With simultaneous consideration of artefacts identification. Then the expert system is to carry on the concluding process on the basis of rules on the correctness of the received input signal in order to use it in the control processes.

Within the frames of works on the concept of the expert system, on the basis of the accessible academic works a framework of knowledge organizing was elaborated. As the knowledge representation method and tool for system construction the MATLAB environment was selected due to the fact of the tool popularity in the environment of persons working on the signal analysis.

The proposed expert system would then be subject to verification and testing.

The general architecture of the system was presented on Fig. 1.

Dissected modules of knowledge gathering and concluding on the basis of the realized expert system were presented on Fig. 2.

While designing the expert system the question of the data implementation was minimized on the basis of heuristics. The academic publications analysis was accepted for the method of knowledge acquisition within the range of the

(6)

EEG signal analysis and the own, confirmed many times academic studies on utilization of the neuroinformatic tools in practice.

Fig. 1. General diagram of expert system

Fig. 2. Dissection of the knowledge gathering and concluding modules of the expert system

To judge the conflicts that may arise during work of the expert interpreting system the following priorities were defined in respect of the used rules: 1.

Importance priority 2. Number of detailed conditions 3. Frequency of use 4.

Time of passing to the knowledge base.

The following rules were used within the frames of elaboration of the expert system concept:

– procedure:

IF <noisy signal> THEN <ICA algorithm>;

– cause-result:

IF <PCA algorithm> THEN <ICA algorithm>;

– declaratory:

IF <eye ball artefact> THEN <artefact elimination algorithm>.

(7)

Also testing of the presented rules was executed on the basis of detection of unnecessary, contradictory, absorbing rules and also the ones with an unnecessary condition or looped. During creation of the expert system the crucial factor is verification of the knowledge base in respect of redundancy, which appears when the unnecessary rules appear. The rules are redundant, if their both conditional parts are fulfilled simultaneously or unfulfilled in all the possible situations. Instead, two rules are redundant, if their conditional parts are fulfilled simultaneously or unfulfilled in all the possible situations and their conclusive parts are different for at least one situation. A rule is absorbed by another one when the conditional part of the first and the second rule is fulfilled and the conclusive parts of the both rules are identical. Then the both rules have unnecessary conditions, if they are both absorbed by the third rule.

5. SUMMARY

The suggested interpreting expert system is based on facts gathered in the knowledge base in the form of patterns of the EEG signal. Thus, uncertainty of the gathered knowledge is low at the same time. Unreliable sources of information were eliminated, so was excessive, irrelevant information and by artefact identification the utmost care was taken so as the unknown factors would not appear influencing the EEG signal.

The advantages of the proposed solution include among others low level of complication of the presentation module and easiness of modification of the facts included in the knowledge base. Thus, the proposed expert system may become a useful tool for constructors of devices or developers of applications based on the brain-computer technology.

REFERENCES

[1] Paszkiel S., The use of Brain Computer Interfaces in the control processes based on industrial PC in terms of the methods of EEG signal analysis, Journal of Medical Informatics & Technologies - Vol.22/2013, pp. 55-62, 2013.

[2] Guger C., Ramoser H., Pfurtscheller G., Real-Time EEG Analysis with Subject- Specific Spatial Patterns for a Brain-Computer Interface, IEEE Transactions on Rehabilitation Engineering, 8(4), pp. 447-456, 2000.

[3] Weiss J.N. et al., Chaos and Chaos Control in Biology, J. Clin. Invest. 93, pp. 1355- 1360, 1994.

[4] Paszkiel S., Augmented reality of technological environment in correlation with brain computer interfaces for control processes, Advances in Intelligent Systems and Computing 267 - AISC, Springer 2014, pp. 197-203, Switzerland 2014.

[5] Rapp P.E. et al., Dynamics of spontaneous neural activity in the simian motor cortex: the dimension of chaotic neurons, Phys. Letters A. 110, pp. 335-338, 1985.

(8)

[6] Chen Y. S., Cheng C. Y., Hsieh J. C., Chen L. F., Maximum contrast beamformer for electromagnetic mapping of brain activity, IEEE Transactions on Biomedical Engineering, 53(9), pp. 1765-74, 2006.

[7] Baillet S., Mosher J. C., Leahy R. M., Electromagnetic brain mapping, IEEE Signal Processing Magazine, 18(6), pp. 14-30, 2001.

[8] Fitzgibbon S. P., Powers D. M., Pope K. J., Clark C. R., Removal of EEG noise and artifact using blind source separation, J Clin Neurophysiol, 24(3), pp. 232-243, 2007.

[9] Paszkiel S., Wykorzystanie metody PCA i ICA do analizy sygnału EEG w kontekście usuwania zakłóceń, Pomiary Automatyka Kontrola, Vol.59, pp. 204-207, Warszawa, 3/2013.

(Received: 20. 01. 2016, revised: 4. 03. 2016)

Cytaty

Powiązane dokumenty

Actions in the field o f employment policy and support for human resources development in Poland are implemented according to the National Measure Plan for

1) The GIS concept of the Atlas of Kraków province was chosen to enable use of data for analysis and decision support. Therefore, the Atlas can be regarded in extension as a

An analysis of the results indicates that different strain and stress in individual regions of the head, including bridging veins, are the consequence of ki-

Despite some limitations, this study presents ex- perimental results collected on in situ sheep brain indentation to determine the mechanical properties of the brain–skull

Key words: brain, biomechanics, finite element method, meshless methods, surgical simulation, image

Именно «отраж ательная, порож даю щ ая (креативная), регулятивно-оценочная и рефлексивная функции» являются важными для

DAY-TO-DAY ORIGIN DESTINATION TUPLE ESTIMATION AND PREDICTION WITH HIERARCHICAL BAYESIAN NETWORKS USING MULTIPLE DATA SOURCES..

[r]