Tackling EEG signal classification with least squares support vector machines: A sensitivity analysis study
Tipo
Artigo
Data de publicação
2010
Periódico
Computers in Biology and Medicine
Citações (Scopus)
56
Autores
Lima C.A.M.
Coelho A.L.V.
Eisencraft M.
Coelho A.L.V.
Eisencraft M.
Orientador
Título da Revista
ISSN da Revista
Título de Volume
Membros da banca
Programa
Resumo
The electroencephalogram (EEG) signal captures the electrical activity of the brain and is an important source of information for studying neurological disorders. The proper analysis of this biological signal plays an important role in the domain of brain-computer interface, which aims at the construction of communication channels between human brain and computers. In this paper, we investigate the application of least squares support vector machines (LS-SVM) to the task of epilepsy diagnosis through automatic EEG signal classification. More specifically, we present a sensitivity analysis study by means of which the performance levels exhibited by standard and least squares SVM classifiers are contrasted, taking into account the setting of the kernel function and of its parameter value. Results of experiments conducted over different types of features extracted from a benchmark EEG signal dataset evidence that the sensitivity profiles of the kernel machines are qualitatively similar, both showing notable performance in terms of accuracy and generalization. In addition, the performance accomplished by optimally configured LS-SVM models is also quantitatively contrasted with that obtained by related approaches for the same dataset. © 2010 Elsevier Ltd.
Descrição
Palavras-chave
Assuntos Scopus
Biological signals , Communication channel , Data sets , EEG signal classification , EEG signals , Electrical activities , Electroencephalogram signals , Human brain , Kernel function , Kernel functions , Kernel machine , Least Square , Least squares support vector machines , Neurological disorders , Parameter values , Performance level , Sensitivity profiles , SVM classifiers , SVM model