Sleep spindle detection using modified extreme learning machine generalized radial basis function method

Combination of discrete wavelet packet transform with detrended fluctuation analysis using customized mother wavelet with the aim of an imagery-motor control interface for an exoskeleton

“…Based on the DWPT-DFA method, the extracted features with different mother wavelets are fed into the SSVM classifier with the GRBF kernel. Regarding the previous studies [8,21,22,30] on the classifiers of NN, K-NN, SVM, RBF and GRBF in the EEG signal processing, we employ a SSVM with a GRBF for classifying the imagin ary movement features. The SSVM and GRBF are the amended algorithms of the traditional SVM and RBF to solve the limitations and add flexibility, such as the curse of dimension in the SVM, and to add flexibility to the RBF for different cases of data distribution [8].…”

“…The SSVM and GRBF are the amended algorithms of the traditional SVM and RBF to solve the limitations and add flexibility, such as the curse of dimension in the SVM, and to add flexibility to the RBF for different cases of data distribution [8]. Based on our previous studies [22,30], the GRBF has the ability of covering the data distribution widely in the feature space, compared to the traditional RBF. The advantage of the SSVM classifier is preventing the algorithm from the curse of dimension by employing the Lagrange theorem, when the number of features is increased.…”

“…The Generalized RBF is a successful kernel, which is utilized in our previous studies [21,22,30]. The flexibility of the GRBF is summarized in terms of width (w > 0), shape (τ > 0) and center (c) of the Gaussian distribution parameters, which leads the fashion to highly accurate and robust results.…”

“…Self-similarity features are also impressively used in the sleep science to detect sleep spindle or k-complex patterns automatically. These patterns are studied for quantifying learning, memory consolidation and detecting mental disorders [22,30]. In our previous study [21], a combination of Katz, Higuchi, and Sevcick fractal dimensions is utilized to quantify the self-similarity, in order to detect sleep spindle patterns.…”

A combination of CSP-based method with soft margin SVM classifier and generalized RBF kernel for imagery-based brain computer interface applications

“…In a classification study based on the BCI, Hoffmann et al [14] showed that the SVM with a RBF kernel has a greater impact on the classification accuracy than other kernels such as polynomial, linear and Sigmoid. In our previous studies [9,[11][12][13]24], the GRBF kernel has higher accuracy and robustness. The procedure of this study is: recording nine EEG data based on the BCI competition III, dataset IVa (the designed task is the same as the BCI Competition III dataset IVa task); implementing the existed DFBCSP-DSLVQ method; applying a frequency sub-band selection method to find the best sub-bands; using the KLDA and KPCA algorithms, in which the kernel is GRBF, to select the features; and finally utilizing different combinations to classify the selected features, namely the SVM with the radial basis function (RBF) kernel (SVM-RBF), SVM with the GRBF kernel (SVM-GRBF), soft margin support vector machine (SSVM) with the RBF kernel (SSVM-RBF), SSVM with the GRBF kernel (SSVM-GRBF), the NN, and the K-nearest neighbor (K-NN).…”