No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
EEG Pattern Recognition Through Multi-Stream Evidence Combination
Abstract
EEG recordings provide an important means of brain-computer communication, but their classification accuracy is limited by unforeseeable variations in the signal due to artefacts or recogniser-subject feedback. A number of techniques were recently developed to address a related problem of recogniser robustness to uncontrollable signal variation which also occurs in automatic speech recognition (ASR). In this article we consider how some of the proved advantages of the "multi-stream combination" and "tandem" approaches in HMM/ANN hybrid based ASR can possibly be applied to improve the performance of EEG recognition. Keywords: EEG, multi-stream classification, robust recognition Acknowledgements: This work was supported by the EC/OFES (European Community / Swiss Federal Office for Education and Science) RESPITE project (REcognition of Speech by Partial Information TEchniques). Contents 1.
Magnetoencephalography (MEG) can be used as an effective non-invasive interface with the brain to provide movement-related information similar to invasive signal recordings. This paper proposes a reliable and efficient algorithm for classification of wrist movement in four directions from MEG signals of two subjects. Our approach involves signal smoothing, design of a class-specific Unique Identifier Signal (UIS) and curve fitting to identify the direction in a given test signal. Our algorithm is evaluated with the data set provided in BCI competition 2008. Our simulations show the best average prediction accuracy of 88.84% for this four-class classification problem. The results of the proposed model are found to be superior to most other techniques in vogue.
ABSTRACT This paper,describes a database,designed,to evaluate the performance,of speech ,recognition ,algorithms ,in noisy conditions.The database may either be used for the evaluation of front-end feature extraction algorithms using a defined HMM recognition,back-end or complete,recognition,systems. The source speech for this database is the TIdigits, consisting of connected,digits task spoken,by American,English talkers (downsampled,to 8kHz) . A selection of 8 different real-world noises have been added to the speech over a range of signal to noise ratiosand,special care has been taken to control the filtering of both the speech and noise. The framework,was,prepared,as a contribution to the ETSI STQ-AURORA DSR Working Group [1] . Aurora is developing standards for Distributed Speech Recognition (DSR) where,the speech analysis is done in the telecommunication terminal and the recognition at a central location in the telecom network. The framework,is currently ,being ,used ,to evaluate ,alternative proposals for front-end feature extraction. The database has been made,publicly available through,ELRA so that other speech researchers can evaluate and compare,the performance,of noise robust algorithms. Recognition results are presented for the first standard DSR feature extraction scheme,that is based on a cepstral analysis.
Electroencephalogram (EEG) recordings during right and left motor imagery can be used to move a cursor to a target on a computer screen. Such an EEG-based brain-computer interface (BCI) can provide a new communication channel to replace an impaired motor function. It can be used by, e.g., patients with amyotrophic lateral sclerosis (ALS) to develop a simple binary response in order to reply to specific questions. Four subjects participated in a series of on-line sessions with an EEG-based cursor control. The EEG was recorded from electrodes overlying sensory-motor areas during left and right motor imagery. The EEG signals were analyzed in subject-specific frequency bands and classified on-line by a neural network. The network output was used as a feedback signal. The on-line error (100%-perfect classification) was between 10.0 and 38.1%. In addition, the single-trial data were also analyzed off-line by using an adaptive autoregressive (AAR) model of order 6. With a linear discriminant analysis the estimated parameters for left and right motor imagery were separated. The error rate obtained varied between 5.8 and 32.8% and was, on average, better than the on-line results. By using the AAR-model for on-line classification an improvement in the error rate can be expected, however, with a classification delay around 1 s.
Proposes a local neural classifier for the recognition of mental
tasks from online spontaneous EEG signals. The classifier is embedded in
a portable brain-computer interface called ABI, which has been evaluated
with 4 young healthy persons. Subjects' performance is analyzed off-line
and, for three of them, also online in the presence of biofeedback. The
proposed ABI recognizes three mental tasks from online spontaneous EEG
signals. Correct recognition is around 70%. This modest rate is largely
compensated by two properties of ABI: wrong responses are below 5% and
it makes decisions every 1/2 second. Also, since the subject and his/her
personal ABI learn simultaneously from each other, subjects master it
rapidly: one of the subjects achieved excellent control in just 5 days
of training
Hidden Markov model speech recognition systems typically use Gaussian mixture models to estimate the distributions of decor- related acoustic feature vectors that correspond to individual sub- word units. By contrast, hybrid connectionist-HMM systems use discriminatively-trained neural networks to estimate the probabil- ity distribution among subword units given the acoustic observa- tions. In this work we show a large improvement in word recog- nition performance by combining neural-net discriminative feature processing with Gaussian-mixture distribution modeling. By train- ing the network to generate the subword probability posteriors, then using transformations of these estimates as the base features for a conventionally-trained Gaussian-mixture based system, we achieve relative error rate reductions of 35% or more on the multi- condition AURORA noisy continuous digits task.
In this paper, we develop different mathematical models in the framework of the multi-stream paradigm for noise robust automatic speech recognition (ASR), and discuss their close relationship with human speech perception. Largely inspired by Fletcher's "product-of-errors'' rule (PoE rule) in psychoacoustics, multi-band ASR aims for robustness to data mismatch through the exploitation of spectral redundancy, while making minimum assumptions about noise type. Previous ASR tests have shown that independent sub-band processing can lead to decreased recognition performance with clean speech. We have overcome this problem by considering every combination of data sub-bands as an independent data stream. After introducing the background to multi-band ASR, we show how this "full combination" approach can be formalised, in the context of hidden Markov model/ artificial neural network (HMM/ANN) based ASR, by introducing a latent variable to specify which data subbands in each data frame are free from data mismatch. This enables us to decompose the posterior probability for each phoneme into a reliability-weighted integral over all possible positions of clean data. This approach offers great potential for adaptation to rapidly changing and unpredictable noise.
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)
This book provides the first comprehensive treatment of feed-forward neural networks from the perspective of statistical pattern recognition. After introducing the basic concepts of pattern recognition, the book describes techniques for modelling probability density functions, and discusses the properties and relative merits of the multi-layer perceptron and radial basis function network models. It also motivates the use of various forms of error functions, and reviews the principal algorithms for error function minimization. As well as providing a detailed discussion of learning and generalization in neural networks, the book also covers the important topics of data processing, feature extraction, and prior knowledge. The book concludes with an extensive treatment of Bayesian techniques and their applications to neural networks.
Nineteen-channel EEG was recorded with closely spaced electrodes overlaying the left sensorimotor cortex during self-paced, voluntary right finger movements. Three right-handed people served as subjects. The EEG was analysed in the 10 Hz band (10-12 Hz) and in four 40 Hz bands (34-36, 36-38, 38-40, 40-42) by calculation of ERD time courses and ERD maps, whereby a ERD is characterized by a movement-related band power decrease. In all three subjects a close to C3 localized 10 Hz ERD was found, starting about 2 s prior to movement onset and continuing during movement. Along with this 10 Hz ERD a localized and short-lasting (about 0.5 s) burst of 40 Hz oscillations was embedded around movement onset. This can be interpreted as indicating that planning of movement is accompanied by a desynchronization of central mu rhythm and a generation of 40 Hz oscillations.
This tutorial provides an overview of the basic theory of hidden
Markov models (HMMs) as originated by L.E. Baum and T. Petrie (1966) and
gives practical details on methods of implementation of the theory along
with a description of selected applications of the theory to distinct
problems in speech recognition. Results from a number of original
sources are combined to provide a single source of acquiring the
background required to pursue further this area of research. The author
first reviews the theory of discrete Markov chains and shows how the
concept of hidden states, where the observation is a probabilistic
function of the state, can be used effectively. The theory is
illustrated with two simple examples, namely coin-tossing, and the
classic balls-in-urns system. Three fundamental problems of HMMs are
noted and several practical techniques for solving these problems are
given. The various types of HMMs that have been studied, including
ergodic as well as left-right models, are described
Automatic speech recognition (ASR) performance falls dramatically with the level of mismatch between training and test data. The human ability to recognise speech when a large proportion of frequencies are dominated by noise has inspired the "missing data" and "multi-band" approaches to noise robust ASR. "Missing data" ASR identifies low SNR spectral data in each data frame and then ignores it. Multi-band ASR trains a separate model for each position of missing data, estimates a reliability weight for each model, then combines model outputs in a weighted sum. A problem with both approaches is that local data reliability estimation is inherently inaccurate and also assumes that all of the training data was clean. In this article we present a model in which adaptive multi-band expert weighting is incorporated naturally into the maximum a posteriori (MAP) decoding process.
Support Vector Machines (SVMs) for regression problems are trained by solving a quadratic optimization problem which needs on the order of l 2 memory and time resources to solve, where l is the number of training examples. In this paper, we propose a decomposition algorithm, SVMTorch 1 , which is similar to SVM-Light proposed by Joachims (1999) for classification problems, but adapted to regression problems. With this algorithm, one can now efficiently solve large-scale regression problems (more than 20000 examples). Comparisons with Nodelib, another publicly available SVM algorithm for large-scale regression problems from Flake and Lawrence (2000) yielded significant time improvements. Finally, based on a recent paper from Lin (2000), we show that a convergence proof exists for our algorithm. 1. Introduction Vapnik (1995) has proposed a method to solve regression problems using support vector machines. It has yielded excellent performance on many regression and time ser...
Design and implementation of an EEG based 'virtual keyboard' using Hidden Markov models
- B Obermaier
Obermaier, B. (2001), "Design and implementation of an EEG based 'virtual keyboard' using Hidden Markov
models", PhD thesis, Technische Universität Graz, Institut für Elektro-und Biomedizische Technik, Abteilung
für Medizinische Informatik, Graz, Austria
Tandem connectionist feature stream extraction for conventional HMM systems Towards ASR on partially corrupted speech
- H Hermansky
- D Ellis
- S Sharma
- H Hermansky
- S Tibrewela
- M Pavel
Hermansky, H., Ellis, D. & Sharma, S. (2000), " Tandem connectionist feature stream extraction for
conventional HMM systems ", ICASSP-2000, http://www.dcs.shef.ac.uk/research/groups/spandh/projects/
respite/publications/ellis_2_icassp_00.pdf
[4]
Hermansky, H., Tibrewela, S. & Pavel, M. (1996) " Towards ASR on partially corrupted speech ", Proc
ICSLP'96, pp. 462-465.
Local neural classifier for EEG-based recognition of mental tasks
- J Millán
- Mourinño
- F Babiloni
- F Cincotti
- M Varsta
- J Heikkonen
Millán, J., Mourinño, Babiloni, F., Cincotti, F., Varsta, M. & Heikkonen, J. (2000), "Local neural classifier for
EEG-based recognition of mental tasks", Proc. IJCNN 2000.