Critical Observations on Interval Type-2 Fuzzy Logic Approach to Uncertainty Handling in a Brain-Computer Interface Design
2006 (English)In: Proc. IPMU 2006, 2006Conference paper (Refereed)
Effective handling of uncertainties associated with variability in brain dynamics and other factors with stochastic characteristics represents a highly challenging problem particularly for existing methods applied to the classification task within a Brain-Computer Interface (BCI). Recently, type-2 fuzzy logic (T2 FL) has been found effective in modelling uncertain data. This paper presents an enhanced Interval T2 FL methodology to the problem of inter-session classification of movement imagination-related patterns in electroencephalogram (EEG) and electrocorticogram (ECoG) recordings. The performance of the devised BCI is assessed based on the classification accuracy (CA) and is found to compare favourably to that of analogous systems employing well-known classical type-1 (T1) FLS and state-of-the-art linear discriminant analysis (LDA) as classifiers. However, the critical issues concerning learning rate selection, rule-base initialisation, selection of optimal model structure, convergence of model parameters and uncertainty bounds initialisation are observed to have a very decisive effect on the robustness of the designed BCI using T2 methodology. The paper presents some practical approaches to effectively tackle some of the issues and highlights the need for further work so that full potential of T2 FLS concept could be exploited.
Place, publisher, year, edition, pages
type-2 fuzzy logic, uncertainty, pattern recognition, electroencephalography, brain-computer interface
IdentifiersURN: urn:nbn:se:kth:diva-90103OAI: oai:DiVA.org:kth-90103DiVA: diva2:504141
11th International Conference on Information Processing and Management of Uncertainty, Paris, France, July 2-7 2006
QC 201205292012-02-192012-02-192012-05-29Bibliographically approved