Vito Logar

Solving the sales prediction problem with fuzzy evolving methods

This paper presents the solving of the petrol sales volume estimation problem given by the Task Force on Competitions, Fuzzy Systems Technical Committee IEEE Computational Intelligence Society. The solution using eTS, SAFIS and eFuMo method is presented. The results are compared to linear ARX model identified using RLS method. Results using static and prediction model are given. The paper also presents parts of new eFuMo method, which is based on recursive Gustafson-Kessel clustering.

Identification of the phase code in an EEG during gripping-force tasks: A possible alternative approach to the development of the brain-computer interfaces

BACKGROUND The subject of brain-computer interfaces (BCIs) represents a vast and still mainly undiscovered land, but perhaps the most interesting part of BCIs is trying to understand the information exchange and coding in the brain itself. According to some recent reports, the phase characteristics of the signals play an important role in the information transfer and coding. The mechanism of phase shifts, regarding the information processing, is also known as the phase coding of information. OBJECTIVE The authors would like to show that electroencephalographic (EEG) signals, measured during the performance of different gripping-force control tasks, carry enough information for the successful prediction of the gripping force, as applied by the subjects, when using a methodology based on the phase demodulation of EEG data. Since the presented methodology is non-invasive it could be used as an alternative approach for the development of BCIs. MATERIALS AND METHODS In order to predict the gripping force from the EEG signals we used a methodology that uses subsequent signal processing methods: simplistic filtering methods, for extracting the appropriate brain rhythm; principal component analysis, for achieving the linear independence and detecting the source of the signal; and the phase-demodulation method, for extracting the phase-coded information about the gripping force. A fuzzy inference system is then used to predict the gripping force from the processed EEG data. RESULTS The proposed methodology has clearly demonstrated that EEG signals carry enough information for a successful prediction of the subjects performance. Moreover, a cross-validation showed that information about the gripping force is encoded in a very similar way between the subjects tested. As for the development of BCIs, considering the computational time to pre-process the data and train the fuzzy model, a real-time online analysis would be possible if the real-time non-causal limitations of the methodology could be overcome. CONCLUSION The study has shown that phase coding in the human brain is a possible mechanism for information coding or transfer during visuo-motor tasks, while the phase-coded content about the gripping forces can be successfully extracted using the phase-demodulation approach. Since the methodology has proven to be appropriate for the case of this study it could also be used as an alternative approach for the development of BCIs for similar tasks.

Gripping-force identification using EEG and phase-demodulation approach.

In this paper we investigate the fuzzy identification of brain-code during simple gripping-force control tasks. Since the synchronized oscillatory activity and the phase dynamics between the brain areas are two important mechanisms in the brains function and information transfer, we decided to examine whether it is possible to extract the encoded information from the EEG signals using the phase-demodulation approach. The EEG was measured during the performance of different visuomotor tasks and the information we were trying to decode was the gripping force as applied by the subjects. The study revealed that it is possible, by using simple beta-rhythm filtering, phase demodulation, principal component analysis and a fuzzy model, to estimate the gripping-force response by using EEG signals as the inputs for the proposed model. The presented study has shown that even though EEG signals represent a superposition of all the active neurons, it is still possible to decode some information about the current activity of the brain centers. Furthermore, the cross-validation showed that the information about the gripping force is encoded in a very similar way for all the examined subjects. Thus, the phase shifts of the EEG signals seem to have a key role during activity and information transfer in the brain, while the phase-demodulation method proved to be a crucial step in the signal processing.

Identification of Human Gripping-force Control from Electro-encephalographic Signals by Artificial Neural Networks

Abstract The exact mechanism of information transfer between different brain regions is still not known. The theory of binding tries to explain how different aspects of perception or motor action combine in the brain to form a unitary experience. The theory presumes that there is no specific center in the brain that would gather the information from all the other brain centers, governing senses, motion, etc., and then make the decision about the action. Instead, the centers bind together, when necessary, maybe through electromagnetic (EM) waves of specific frequency. Therefore, it is reasonable to assume that the information that is transferred between the brain centers is somehow coded in the electro-encephalographic (EEG) signals. The aim of this study was to explore whether it is possible to extract the information on brain activity from the EEG signals during visuomotor tracking task. In order to achieve the goal, artificial neural network (ANN) was used to predict the measured gripping-force from the EEG signal measurements and thus to show the correlation between EEG signals and motor activity. The ANN was first trained with raw EEG signals of all the measured electrodes as inputs and gripping-force as the output. However, the ANN could not be trained to perform the task successfully. If we presume that brain centers transmit and receive information through EM signals, as suggested by the binding theory, a simplified model of signal transmission in brain can be proposed. We propose a mathematical model of a human brain where the information between centers is transmitted as phase-modulated signal of certain carrier frequency. Demodulated signals were then used as the inputs for the ANN and the gripping-force signal was estimated on the output. The ANN could be trained to efficiently predict the gripping-force signal from the phase-demodulated EEG signals.

Motivation Experiments for Complex Control Systems Education

Abstract New organization scheme according to Bologna study at the Faculty of Electrical Engineering, University of Ljubljana has resulted in modification of lectures also at the Department of Automatic Control. In the paper organization of education process is presented for the subject entitled “Advanced control design methods” at the level of master study. For motivation purposes the subsequent main goals were followed: transparent connections between theory and practice, flexibility in studying possibilities, excellent final results in knowledge presentation and introduction into research activities. To fulfil desired aims laboratory exercises and exams are combined into students’ projects which can be realized using personal computers, laboratory pilot plants and corresponding equipment and partly also through e-learning platform E-CHO enabling among other functionality also virtual and remote experimentations. The final students’ work evaluation includes also corresponding oral presentations where the results have to be defended.This paper aims to obtain adaptive controllers capable of interacting and controlling a physical system in real time. This task is accomplished using a model with a DC motor platform in order to carry out test and to obtain experimental results. It could then be possible to extend these controllers to other systems with minor changes. To achieve this purpose, MATLAB and Simulink simulation environment are used, creating the models needed through block languages.

Visuo-Motor Tasks in a Brain-Computer Interface Analysis

The neurophysiological studies covered by the subject of Brain-Computer Interfaces (BCIs) (del R. Millan et al., 2004; Lebedev & Nicolelis, 2006; Wolpaw et al., 2002) represent a promising, but so far rather undiscovered, area of research. What is perhaps the most interesting part of BCI research is the idea of understanding the information coding in the brain and its use when performing different predefined actions or commands. Recent reports have proposed various techniques for the development of BCIs, based either on the electroencephalographic (EEG) non-invasive (Birbaumer et al., 1999; Wolpaw & McFarland, 2004), invasive (Taylor et al., 2002; Wessberg et al., 2000), magnetoencephalographic (MEG) (Georgopoulos et al., 2005; Mellinger et al., 2007) or other (fMRI, PET, optical imaging) measurements. Since all of these, except EEG, still represent technically demanding and expensive methods, the EEG-based BCIs tend to prevail. Modern BCIs are often classified into several groups based on the electrophysiological signals used, i.e., the different brain potentials (evoked visual, slow cortical, P300 evoked), the mu and beta rhythms, the activities of single cortical neurons, etc. (Wolpaw et al., 2002). The human brain can be considered as a system of highly interconnected groups of neurons, where each neuron or group acts as an oscillator. When the brain is in a certain mode or state, different groups of these neurons synchronize themselves to a certain physiological frequency. In order to achieve a large-scale neuronal synchronization that is detectable, for instance, when using an EEG, several tens of thousands of neurons need to fire at approximately the same time with respect to a neuronal population that has approximately the same spatial orientation. It is believed that the theory of oscillations represents one of the essential mechanisms of brain operation, as studies have shown that every single process in the brain is probably within the neuronal system mediated by means of the electric oscillations of the neuronal populations (Engel et al., 2001). These oscillations or oscillatory activity can be classified into different frequency bands and are referred to as the brain rhythms ([0.5 − 3Hz] – delta, [4 − 7HZ] – theta, [8 − 12Hz] – alpha, [13 − 30Hz] – beta, [30 − 50Hz] – gamma). It is suggested that the synchronization of the oscillatory activity carries out the brain’s functionality, cognition and behavior, which are based on distributed, parallel information processing and exchange between anatomically not necessarily connected neuronal populations (Ivanitsky et al., 2001; Manganotti et al., 1998; Pfurtscheller & Andrew, 1999). When a collaboration of neuronal populations is necessary to 7

Remote laboratory for e-learning of multivariable control design

Regarding technical development different configurations and organizations of e-learning are becoming more and more popular. In spite of the fact that it is frequently used at different educational levels there is always a challenge how to combine in a suitable manner remote learning with the classical one to take advantage of both possibilities. In the paper a strategy is described which was tested through the environment of E-CHO platform in one - semester course regarding Multivariable systems control design at the Faculty of electrical engineering in Ljubljana, Slovenia.

STUDENT COMPETITION IN MULTIVARIABLE CONTROL DESIGN - EDUCATION THROUGH E-GAME

In the paper some ideas of studying multivariable control are presented where special attention is devoted to step - by step transition to e-learning. Introduced ideas are realized through design projects between which one is chosen as a competition game and is realized using E-CHO system, Matlab and a pilot plant. All preparation for the game can be realized using remote virtual and actual laboratory device while the competition itself is taking place in the laboratory to preserve personal contacts between the students and the staff. Copyright

Modelling and Simulation Experimentation through E-CHO Learning Environment

In the paper E-CHO e-learning environment and its main characteristics are presented and analysed. This program was developed at the Faculty of Electrical Engineering, University of Ljubljana, Slovenia, where cooperation between different laboratories enables integration of characteristics important for education purposes and specific for different research areas. Some of them are illustrated through a modelling and simulation examples which enable also several further extensions. These examples are divided into the sequence of steps or sequences guiding students through important dynamic system properties, modelling and simulation, and evaluation of obtained results with respect to experimentation measurements. Built-in flexibility enables also the usage of virtual and real remote experimentation as well as different repetition and testing possibilities.

A possible use of EEG signals in a brain-computer interface.

Background: Newest insights in the field of brain-information coding suggest that the information is transferred between the active regions of the brain as a phase-coded content. Considering the informational richness of the electroencephalographic (EEG) signals, we can assume that by using appropriate methods of signal processing it is possible to decode some of this information. The authors would like to show that using a phase-demodulation approach it is possible to successfully decode the information about the wrist movements of a complex dynamic visuo-motor task (dVM). Since the causality of the methodology is assured, it is also usable for the development of a brain-computer interface (BCI). Methods: In this study we measured the EEG signals from four subjects while performing a dynamic visuo-motor task. For decoding the information, which is supposedly carried by the EEG signals we used brain-rhythm filtering, phase demodulation and principal component analysis approach. As a prediction model for wrist movements, fuzzy inference model was used. Results: The presented results show that the EEG signals measured during the performance of dVM tasks carry enough information about the current action for satisfactory decoding and prediction of the wrist movements. Successful estimation of the motor action is proved also by obtaining reasonably high values of the correlation coefficients. Conclusions: The study has shown that using the proposed methodology it is possible to decode the EEG information of the wrist movements during dVM tasks. The study has also shown that these relatively simple methods of signal processing and a fuzzy model are applicable to the development of a closed-loop, non-invasive BCI.