No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Hidden Markov models used for the offline classification of EEG data
Abstract
Hidden Markov models (HMM) are introduced for the offline classification of single-trail EEG data in a brain-computer-interface (BCI). The HMMs are used to classify Hjorth parameters calculated from bipolar EEG data, recorded during the imagination of a left or right hand movement. The effects of different types of HMMs on the recognition rate are discussed. Furthermore a comparison of the results achieved with the linear discriminant (LD) and the HMM, is presented.
... A few teams have investigated the interest of taking into account the sequential nature of the neural features or of the class labels ( Obermaier et al., 2001;Chiappa and Bengio, 2003;Argun?sahArgun?sah and ?etin, 2010). ...
... An alternative approach has been investigated in offline preliminary studies (Obermaier et al., 2001;Darmanjian et al., 2003;Argun?sahArgun?sah and ?etin, 2010;Onaran et al., 2011;Wissel et al., 2013). ...
... in EEG signals ( Obermaier et al., 2001;Argun?sahArgun?sah and ?etin, 2010). ...
Brain-Computer Interfaces (BCIs) are systems that establish a direct communication pathway between the users' brain activity and external effectors. They offer the potential to improve the quality of life of motor-impaired patients. Motor BCIs aim to permit severely motor-impaired users to regain limb mobility by controlling orthoses or prostheses. In particular, motor BCI systems benefit patients if the decoded actions reflect the users' intentions with an accuracy that enables them to efficiently interact with their environment. One of the main challenges of BCI systems is to adapt the BCI's signal translation blocks to the user to reach a high decoding accuracy. This paper will review the literature of data-driven and user-specific transducer design and identification approaches and it focuses on internally-paced motor BCIs. In particular, continuous kinematic biomimetic and mental-task decoders are reviewed. Furthermore, static and dynamic decoding approaches, linear and non-linear decoding, offline and real-time identification algorithms are considered. The current progress and challenges related to the design of clinical-compatible motor BCI transducers are additionally discussed.
... The classification results reached with the HMM approach are better than those reached with neural networks; see [2]. Compared with other existing systems ([2], [11], [12], [14], [15], [17]) this contribution tries to classify the movements related to one side of the body. ...
... The classification results reached with the HMM approach are better than those reached with neural networks; see [2]. Compared with other existing systems ([2], [11], [12], [14], [15], [17]) this contribution tries to classify the movements related to one side of the body. This task is much more complicated. ...
... Hence no conversion was used anymore. Also that's why only one model mixture was used (in contrast to [2]). For details of the experiments performed see [3]. ...
The contribution describes the design, optimization and verificationof the off-line single-trial movement classification system. Four typesof movements are used for the classification: the right index fingerextension vs. flexion as well as the right shoulder (proximal) vs.right index finger (distal) movement. The classification systemutilizes hidden information stored in the characteristic shapes ofhuman brain activity (EEG signal). The great variability of EEGpotentials requires using of context information and hence theclassifier based on Hidden Markov Models (HMM). The suitableparameterization, model structure as well as training andclassification process are suggested on the base of spectral analysisresults and experience with the speech recognition. The training andthe classification are performed with the disjoint sets of EEGrealizations. Classification experiments are performed with 10 randomlychosen sets of EEG realizations. The final average score of thedistal/proximal movement classification is 80%; the standard deviationof classification results is 9%. The classification of the extension /flexion gives comparable results.
... The majority of research on BCI has been based on electroencephalography (EEG) data and restricted to simple experimental tasks using a small set of commands. In these studies (Cincotti et al 2003, Lee and Choi 2003, Obermaier et al 1999) information was extracted from a limited number of EEG channels over scalp sites of the right and left hemisphere. ...
... However, depending on the individual problem, distinct properties of other machine learning methods may provide another viable BCI approach. Pascual-Marqui et al (1995) and Obermaier et al (1999) argued that brain states and state transitions can explain components of observed human brain activity. Obermaier et al applied HMMs in offline classification of non-invasive EEG recordings of brain activity to distinguish between two different imagined movements. ...
Objective. Support vector machines (SVM) have developed into a gold standard for accurate classification in brain–computer interfaces (BCI). The choice of the most appropriate classifier for a particular application depends on several characteristics in addition to decoding accuracy. Here we investigate the implementation of hidden Markov models (HMM) for online BCIs and discuss strategies to improve their performance. Approach. We compare the SVM, serving as a reference, and HMMs for classifying discrete finger movements obtained from electrocorticograms of four subjects performing a finger tapping experiment. The classifier decisions are based on a subset of low-frequency time domain and high gamma oscillation features. Main results. We show that decoding optimization between the two approaches is due to the way features are extracted and selected and less dependent on the classifier. An additional gain in HMM performance of up to 6% was obtained by introducing model constraints. Comparable accuracies of up to 90% were achieved with both SVM and HMM with the high gamma cortical response providing the most important decoding information for both techniques. Significance. We discuss technical HMM characteristics and adaptations in the context of the presented data as well as for general BCI applications. Our findings suggest that HMMs and their characteristics are promising for efficient online BCIs.
... A linear classi®er like the LD can suciently be the limited amount of available training data ± training of the non-linear HMM-based classi®er based on the same amount of data seemed critical therefore the need of a dimension reduction. Oine analysis of BCI data recorded during earlier sessions revealed that HMM in combination with Hjorth parameters are suitable to classify EEG signals related to imagination of either a left or right hand movement (Obermaier et al., 1999). Preliminary results of online classi®cation of EEG patterns in an HMMbased system were presented in (Obermaier et al., 2000). ...
... The number of states ranged from 1 to 5, which corresponds to physiological changes in the spatio-temporal patterns in a 1 s range (Schloegl et al., 1997a). The number of mixtures was limited to eight referring to earlier studies made by the authors (Obermaier et al., 1999). ...
Hidden Markov models (HMMs) are presented for the online classification of single trial EEG data during imagination of a left or right hand movement. The classification shows an improvement of the online experiment and the temporal determination of minimal classification error compared to linear classification methods.
... Frequency components in these bands are predominately involved in motor imagery [14]–[16]. The calculation was performed sample by sample in the time domain using a fifth-order Butterworth filter in a window from seconds 4 to 8 of each trial (see [12]). Therefore, the feature vector describing all EEG signals from all electrodes had 145 components. ...
... The maximum number of states was limited to five, which corresponds to physiological changes in the spatiotemporal patterns in a one second range [21]. The number of mixtures was limited to eight, according to earlier studies made by the authors [12]. The Gaussian mixtures were approximated based on a -means clustering of the feature vectors, whereas the number of mixtures corresponds to the number of estimated clusters. ...
The information transfer rate, given in bits per trial, is used as an evaluation measurement in a brain-computer interface (BCI). Three subjects performed four motor-imagery (left hand, right hand, foot, and tongue) and one mental-calculation task. Classification of the electroencephalogram (EEG) patterns is based on band power estimates and hidden Markov models (HMMs). We propose a method that combines the EEG patterns based on separability into subsets of two, three, four, and five mental tasks. The information transfer rates of the BCI systems comprised of these subsets are reported. The achieved information transfer rates vary from 0.42 to 0.81 bits per trial and reveal that the upper limit of different mental tasks for a BCI system is three. In each subject, different combinations of three tasks resulted in the best performance.
... One of these relatively novel methods is Hidden Markov Modelling (HMM), an unsupervised machine learning technique that reconstructs a sequence of patterns as a system of temporally-discrete states. Previously, HMMs have been used to extract the underlying dynamical properties of neural data from MEG (Baker et al., 2014;Vidaurre et al., 2018b;Quinn et al., 2018, Hawkins et al., 2019, EEG (Obermaier et al., 2001;Williams et al., 2018;Dash & Kolekar, 2020;Marzetti, 2023), and fMRI (Duan et al., 2005;Dang et al., 2017;Goucher-Lambert & McComb, 2019;Hussain et al., 2023) at rest and in task settings. To our knowledge, there have not been any studies of how resting-state neural dynamics vary in a developmental sample, or how these dynamics relate to the emergence of diverse profiles of behaviour and cognitive ability. ...
Resting-state network activity has been associated with the emergence of individual differences across childhood development. However, due to the limitations of time-averaged representations of neural activity, little is known about how cognitive and behavioural variability relates to the rapid spatiotemporal dynamics of these networks. Magnetoencephalography (MEG), which records neural activity at a millisecond timescale, can be combined with Hidden Markov Modelling (HMM) to track the spatial and temporal characteristics of transient neural states. We applied HMMs to resting-state MEG data from (n = 46) children aged 8-13, who were also assessed on their cognitive ability and across multiple parent-report measures of behaviour. We found that entropy-related properties of participants’ resting-state time-courses were positively associated with cognitive ability. Additionally, cognitive ability was positively correlated with the probability of transitioning into HMM states involving fronto-parietal and somatomotor activation, and negatively associated with a state distinguished by default-mode network suppression. We discuss how using dynamical measures to characterise rapid, spontaneous patterns of brain activity can shed new light on neurodevelopmental processes implicated in the emergence of cognitive differences in childhood.
Significance Statement
There is increasing evidence that the function of resting-state brain networks contributes to individual differences in cognition and behaviour across development. However, the relationship between dynamic, transient patterns of switching between resting-state networks and neurodevelopmental diversity is largely unknown. Here, we show that cognitive ability in childhood is related to the complexity of resting-state brain dynamics. Additionally, we demonstrate that the probability of transitioning into and remaining in certain ‘states’ of brain network activity predicts individual differences in cognitive ability.
... Hidden Markov Model has been used by Obermaier et al., [39,40]. An algorithm based on a HMM has been showcased which explicitly models the ERD and ERS features to classify real finger movements (flexion and extension) [41]. ...
In the past few decades research groups and scientific establishments all over the world
have delved deeper into the understanding of psychomotor symptoms of various disorders like
depressions and some neuromotor disorders. For systematic study of the symptoms of these
disorders and finding effective there is a need of effective communication between Human
Brain and the external computer or device. In this paper a data a low-cost point of care testing
system is developed from the scratch. The Data Acquisition part of this system is implemented
with the help of custom made electrodes and NI’s DAQ USB 6008. The NI DAQ has an inbuilt
ADC which helps digitize the acquired analog signals. For creating and implementing the
algorithm, LabVIEW visual programing language is used. All of the preprocessing and
postprocessing of the acquired signal ia done in LabVIEW using some customizable functions
called VI’s. The acquired data is displayed to the user in real time in the LabVIEW with the help
of waveform graphs, text indicators and LED’s. The system is thoroughly jitter tested for
checking its robustness and minimal data samples loss is achieved. While testing, optimum data
acquisition parameters are also endeavoured to be determined and the analysed data of testing is
presented in Tables and plots. The optimum parameters determinded for acquisition are found to
be achieveing a high level of accuracy in data acquisition process. Further, a simple EEG data
acquisition task is performed on one subject by recording the EEG data coming during eyes open
and closed conditions both when subject is sitting and when the subject is standing. The results
of the analysed data have shown that the single ended electrode mode for this system is more
accurate than the differential electrode mode. Moreover, the O1 O2 electrodes placement
provided more comprehensible data in comparison to the data provided by Cz O1 electrodes
placement. A classifier was tried and fitted (Threshold detector) to discriminate the signals into
two states i.e. eyes open/close and the output state feedback is given via a GUI based LED
named Brain Switch.
... We have read with great interest the recent paper by Gärtner et al. (2015) on EEG microstate sequences. These authors propose a stochastic model termed sampled marked intervals (SMI), relating the observed microstate sequence to an assumed underlying stochastic process, similar to earlier work that proposed Hidden Markov Models (HMM) (De Lucia et al., 2011;Obermaier et al., 1999). We very much appreciate such a theoretical approach to enrich the classical EEG microstate analysis (Lehmann and Skrandies, 1984). ...
Highlights
• EEG microstate sequences of the resting brain exhibit long-range dependency (LRD).
• We describe why LRD is important and provide comparisons to other complex systems.
• Hidden Markov Models (HMM) do not normally capture LRD.
• We propose to test whether the proposed model of Gärtner et al. captures LRD.
... For example, galvanic skin response (GSR) which measures the conductivity of the skin is used to see the correlation of sweat glands which respond to the psychological stimulation [36]. GSR also has been used as an indicator of experience in the study of non-technical domains [42] and technical domains [43] On the other hand, other researchers choose other emotional physiological processing methods in the study of emotion include Electrocardiograms (ECG), which compute heart rate, inter-beat interval, heart rate variability and blood pressure [44], Electromyography (EMG) which measures muscle activity [9], Electrooculography (EOG) which measures the resting potential of the retina and recoding the movement of the eyes [45] and Electroencephalogram (EEG) which detects brain electrical activity [4] [46]. Table 2 summarizes the use of the physiological signals in the area of HCI, with the key attributes and the bio-signals selection. ...
The issue of study in emotion or affect has been recently examined by Human-computer Interaction (HCI) research groups, in particular for the development of affective interaction and design. With the recent technological advances, humans are able to interact with computers in ways which are almost impossible. The new modalities for computer interaction with human emotion such skin conductivity, heart rate, brain signals and physiological signals are emerging, which extend beyond the confines of the keyboard and mouse. As such, emotion plays an important role in human-to-human communication and interaction, which allows people to express it beyond the verbal domain. The physiological signals, which are the focus of this study, play a significant method for usability study in HCI. Numbers of usability psychometrics are developed, however, a few yield "natural" interpretations, such as feelings. The physiological signals can be interpreted as a trustworthy of emotional reveal. In this study, we conduct an affective interaction and design development for an e-Learning storyboard tool called SCOUT. The paper describes the significant role of physiological signals as a method to study usability in emotion in the area of HCI and Affective Computing. Affective design and interaction are also discussed in the SCOUT development. The comparison of different research areas in HCI, focusing on different key attributes and physiological signals are studied. The result of the analysis revealed that physiological signal should be used as important as other usability psychometric in HCl, to enrich the evaluation of human-computer interaction study with human emotion.
... GSR also has been used as an indicator of experience in the study of non-technical domains [7] and technical domains [31] [34]. On the other hand, other researchers choose other physiological metrics methods in the study of emotion include Electrocardiograms (ECG), which compute heart rate, inter-beat interval, heart rate variability and blood pressure [27], Electromyography (EMG) which measures muscle activity [9], Electrooculography (EOG) which measures the resting potential of the retina and recoding the movement of the eyes [12] and Electroencephalogram (EEG) which detects brain electrical activity [2] [6]. Besides of the advantages of using physiological metrics, there are some disadvantages. ...
Human-Computer Interaction (HCI) research groups have recently attracted to the issue of emotion or affect especially in the examination of interaction and design. With recent technological advances, human users are able to interact with computers in ways which are almost impossible. New modalities for computer interaction with human emotion such as skin conductivity, heart rate, brain signals and physiological signals are emerging. It shows that emotion plays an important role in human communication and interaction, therefore allow people to express emotion beyond the verbal domain. This issue motivates the investigation of two modals of emotional processing in the application of HCI and User Interface Design (UID) areas. The result of this study is directed to the development of an affective interaction design storyboard tool called SCOUT. The paper addresses significant roles of Multi-modal Emotional Processing methods for SCOUT, which includes different types of Psychometric usability methods and Physiological emotional processing methods. The application of Psychometrics and Multi-modal Emotional Processing are then, analyzed. The results of the analysis revealed that the use of both processing methods would enrich the evaluation of emotion in human-computer interaction study.
... HMMs were first successfully applied for speech recognition, and later in molecular biology for modeling the probabilistic profile of protein families (Rabiner, 1989). HMM has been successfully used in a BCI application for online classification of EEG signals acquired during left-hand and right-hand motor imagery (Obermaier et al., 1999). To our knowledge, this is the first time that SVM and HMM techniques have been used to classify NIRS signals for the development of a BCI. ...
There has been an increase in research interest for brain-computer interface (BCI) technology as an alternate mode of communication and environmental control for the disabled, such as patients suffering from amyotrophic lateral sclerosis (ALS), brainstem stroke and spinal cord injury. Disabled patients with appropriate physical care and cognitive ability to communicate with their social environment continue to live with a reasonable quality of life over extended periods of time. Near-infrared spectroscopy is a non-invasive technique which utilizes light in the near-infrared range (700 to 1000 nm) to determine cerebral oxygenation, blood flow and metabolic status of localized regions of the brain. In this paper, we describe a study conducted to test the feasibility of using multichannel NIRS in the development of a BCI. We used a continuous wave 20-channel NIRS system over the motor cortex of 5 healthy volunteers to measure oxygenated and deoxygenated hemoglobin changes during left-hand and right-hand motor imagery. We present results of signal analysis indicating that there exist distinct patterns of hemodynamic responses which could be utilized in a pattern classifier towards developing a BCI. We applied two different pattern recognition algorithms separately, Support Vector Machines (SVM) and Hidden Markov Model (HMM), to classify the data offline. SVM classified left-hand imagery from right-hand imagery with an average accuracy of 73% for all volunteers, while HMM performed better with an average accuracy of 89%. Our results indicate potential application of NIRS in the development of BCIs. We also discuss here future extension of our system to develop a word speller application based on a cursor control paradigm incorporating online pattern classification of single-trial NIRS data.
Hidden Markov Models (HMMs) have emerged as a powerful tool for analyzing time series of neural activity. Gaussian HMMs and their time-resolved extension, Time-Delay Embedded HMMs (TDE-HMMs), have been instrumental in detecting discrete brain states in the form of temporal sequences of large-scale brain networks. To assess the performance of Gaussian HMMs and TDE-HMMs in this context, we conducted simulations that generated synthetic data representing multiple phase-coupled interactions between different cortical regions to mimic real neural data. Our study demonstrates that TDE-HMM performs better than Gaussian HMM in accurately detecting brain states from synthetic phase-coupled interaction data. Finally, for TDE-HMMs, we manipulated key parameters such as phase coupling variability, state duration, and influence of volume conduction effect to evaluate the models' performance under varying conditions.
EEG signal can be a good alternative for disabled persons who cannot perform actions or perform them improperly. Brain computer interface (BCI) is an attractive technology which permits control and interaction with a computer or a machine using EEG signals. Brain task identification based on EEG signals is very difficult task and is still challenging researchers. In this paper, the motor imagery of left and right hand actions are identified using new features which are fed to a set of optimized SVM classifiers. Multi classifiers based classification showed having high faculty to improve the classification accuracy when using different kind or diversified features. Features selection was performed by genetic algorithm optimization. In single optimized SVM classifier, a mean classification accuracy of 89.8% was reached. To further improve the rate of classification, three SVMs classifiers have been suggested and optimized in order to find suitable features for each classifier. The three SVMs classifiers were optimized and achieved a performance mean of 94.11%. The achieved performance is a significant improvement comparing to the existing methods which does not exceed 81% while using the same database. Here, combining multi classifiers with selecting suitable features by optimization can be a good alternative for BCI applications.
The chapter provides an over view of some of the major feature extraction, pattern recognition, and machine learning techniques that have been successfully applied to electroencephalography (EEG)- and ECoG-based Brain-computer interfaces (BCIs). BCIs augment the human ability to communicate and interact with the external world by directly linking the brain to computers and robotic devices. BCIs bypass the normal neuromuscular output pathways for translating brain signals into action. Instead, physiological brain signals are processed in real time by digital signal processing methods to allow a novel form of communication and interaction with the environment. A major goal of BCIs has been to improve the quality of life of physically impaired individuals, including those paralyzed because of degenerative neurological diseases such as amyotrophic lateral sclerosis, spinal cord injury, or stroke. BCIs allow a subject to directly control objects such as a cursor or a robot using brain signals. BCIs depend crucially on the ability to reliably identify behaviorally induced changes (or "cognitive states") in the brain signals being recorded. Statistical pattern recognition and machine learning algorithms play an important role in identifying these changes in brain signals and mapping them to appropriate control signals.
We report about an automatic continuous sleep stager which is based on probabilistic principles employing Hidden Markov Models (HMM). Our sleep stager offers the advantage of being objective by not relying on human scorers, having much finer temporal resolution (one second instead of 30 seconds),and being based on solid probabilistic principles rather than a predefined set of rules (Rechtschaffen and Kales). Results obtained for nine whole night sleep recordings are reported.
Changes in EEG power spectra related to the imagination of movements may be used to build up a direct communication channel between brain and computer (brain computer interface; BCI). However, for the practical implementation of a BCI device, the feature classifier plays a crucial role. We compared the performance of three different feature classifiers for the detection of the imagined movements in a group of 6 normal subjects by means the EEG. The feature classifiers compared were those based on the hidden Markov models (HMM), the artificial neural network (ANN) and on the Mahalanobis distance (MD). Results show a better performance of the MD and ANN classifiers with respect to the HMM classifier.
The article describes the classification of simple movements using a system based on Hidden Markov Models (HMM). Brisk extensions and flexions of the index finger, and movements of the proximal arm (shoulder) and distal arm (finger) were classified using scalp EEG signals. The aim of our study was to develop a system for the classification of movements which show EEG changes at identical scalp electrodes of one hemisphere. The classification of EEG patterns related to movements of one hand is difficult because the disentanglement of movements can only rely on the temporal evolution of EEG changes at one recording site. A large variability-of EEG waveforms requires the use of the context information. The classification procedure was optimized in all parts to increase the recognition score and it was extensively tested on a set of EEG data. The average classification score was 80%, std. deviation 9% for the classification of distal and proximal movements. The classification of extension/flexion reached even better results (due to more accurate localization of the signal source on the scalp). The classification of movement-related EEG data based on HMM yielded higher recognition scores than previously reported classification scores based on artificial neural networks.
Knowing that the driving task of a conventional wheelchair could be difficult or even impossible for impairment people, this work presents an overview of some strategies developed to aid these people. Within this context, a myoelectrical eye-blink and an iris tracking system to guide a robotic wheelchair are briefly described. Futhermore, some comments about EEG-based systems are also presented. Finally, it is presented a robotic wheelchair navigation system capable to reach a desired pose in a planar environment while avoiding static and dynamic obstacles.
Classification of the electroencephalogram (EEG) during motor
imagery of the left or right hand can be performed using a classifier
comprising two hidden Markov models (HMMs) describing the
spatio-temporal patterns related to the imagination. Due to the known
asymmetries during motor imagery of rightand left-hand movement, an
HMM-based classifier allowing asymmetrical structures is introduced. The
comparison between such a system and a symmetrical one is based on the
error rate of classification. The results for EEG data collected during
20 sessions from five subjects demonstrate a significant improvement of
9% for the classification accuracy for the asymmetric classifiers. The
selection of the DAM for classification is done using a variant of
genetic algorithms (GAs); namely, the adaptive reservoir genetic
algorithm (ARGA)
About the Series: Bioelectric Engineering presents state-of-the-art discussions on modern biomedical engineering with respect to applications of electrical engineering and information technology in biomedicine. This focus affirms Springer's commitment to publishing important reviews of the broadest interest to biomedical engineers, bioengineers, and their colleagues in affiliated disciplines. Recent volumes have covered modeling and imaging of bioelectric activity, neural engineering, biosignal processing, bionanotechnology, among other topics.
Thesis (Ph. D.)--University of California, San Diego, 2003. Vita. Includes bibliographical references (leaves 299-322).
I have always been infatuated with computers and convinced of their potential for solving problems in biologic research and clinical care. In the 1960s I thought we could use the computer to predict the shape of macromolecules from their chemical formulas and fundamental physical chemical principles. However, with the computers of the 1960s that was a fantasy. So I focused on the use of computers to manage medical record content and to assist with clinical care. The Electronic Medical Record (EMR) we began developing in 1972 with 33 diabetes patients now carries nearly 300 million separate results for more than 3 million patients. The data include lab and other diagnostic studies, dictated notes, orders, encounter records, radiology images, electrocardiograph tracings, and motion cardiac echoes, and the care provider at Indiana University and Wishard Hospital is accessed 10 million times per year. We have also agitated for standards to make the collection of these data easier. This work has become part of a field called medical informatics. In the meantime, the application of computers to biology has rapidly matured into a field called bioinformatics, and researchers in this field now provide annotated databases for many categories of molecules, programs for "matching" newly discovered genomic sequences with previously studied sequences, and systems for storing and processing massive amounts of genomic and molemic data. They have developed sophisticated methods for predicting the shape of biologic macromolecules and other important insights about biology and evolution. Medical informatics and bioinformatics intersect at many points. The most important intersection is between electronic medical records and the human specimen databases that can link genotype to the phenotype, as needed, to unravel polygenetic disease causality. The National Cancer Institute is embarking on an intriguing effort to use EMRs (phenotype) to link to paraffin blocks (genotype) in pathology laboratories where opportunities for cancer genomic discovery are open. We will participate in this effort and look forward to bending the EMR we developed for clinical use to bioinformatics uses as well.
The ultrasound (US) scattered signal from blood has been treated as a random signal by many investigators. However, the degree of randomness of a medium is a relative term that can change considerably with the resolution of the sensor. In this study, the backscattered signal from blood has been looked at as a chaotic signal. By this treatment, according to Taken's theorem, a single variable (e.g., amplitude of the blood-backscattered signal) can be used to reconstruct the nonlinear dynamics of the blood-scattered signal. Multilayer perceptron neural network architecture, with error back-propagation, has been formulated and used as a basis for building and testing the chaotic model of the backscattered signal. This chaotic model is used successfully as a short-term predictor of the backscattered signal from blood-mimicking fluid (BMF) flowing in a vascular flow phantom under pulsatile flow. This modelling approach can be useful, for example, in detecting blood-borne emboli.
To reduce the amount of computation in a full search (FS)
algorithm for fast motion estimation, we propose a novel and fast FS
motion estimation algorithm. The computational reduction without any
degradation in the predicted image comes from fast elimination of
impossible motion vectors. We obtain faster elimination of inappropriate
motion vectors using efficient matching units from localization of the
complex area in image data. We show three properties in block matching
of motion estimation. We suggest two fast matching scan algorithms, one
from adaptive matching scan and the other from fixed dithering order.
Experimentally, we remove the unnecessary computations by about 30% with
our proposed algorithm compared to conventional fast matching scan
algorithms
EEG classification using Learning Vector Quantization (LVQ) is introduced on the basis of a Brain-Computer Interface (BCI) built in Graz, where a subject controlled a cursor in one dimension on a monitor using potentials recorded from the intact scalp. The method of classification with LVQ is described in detail along with first results on a subject who participated in four on-line cursor control sessions. Using this data, extensive off-line experiments were performed to show the influence of the various parameters of the classifier and the extracted features of the EEG on the classification results.
The paper describes work on the brain--computer interface (BCI). The BCI is designed to help patients with severe motor impairment (e.g. amyotropic lateral sclerosis) to communicate with their environment through wilful modification of their EEG. To establish such a communication channel, two major prerequisites have to be fulfilled: features that reliably describe several distinctive brain states have to be available, and these features must be classified on-line, i.e. on a single-trial basis. The prototype Graz BCI II, which is based on the distinction of three different types of EEG pattern, is described, and results of online and offline classification performance of four subjects are reported. The online results suggest that, in the best case, a classification accuracy of about 60% is reached after only three training sessions. The online results show how selection of specific frequency bands influences the classification performance in single-trial data.
The study focuses on the problems of dimensionality reduction by means of principal component analysis (PCA) in the context of single-trial EEG data classification (i.e. discriminating between imagined left- and right-hand movement). The principal components with the highest variance, however, do not necessarily carry the greatest information to enable a discrimination between classes. An EEG data set is presented where principal components with high variance cannot be used for discrimination. In addition, a method based on linear discriminant analysis (LDA), is introduced that detects principal components which can be used for discrimination, leading to data sets of reduced dimensionality but similar classification accuracy.