Epileptic Seizure Detection Using EEGs Based on Kernel Radius of Intrinsic Mode Functions
The study of automated epileptic seizure detection using EEGs has attracted more and more researchers in these decades. How to extract appropriate features in EEGs, which can be applied to differentiate non-seizure EEG from seizure EEG, is considered to be crucial in the successful realization. In this work, we proposed a novel kernel-radius-based feature extraction method from the perspective of nonlinear dynamics analysis. The given EEG signal is first decomposed into different numbers of intrinsic mode functions (IMFs) adaptively by using empirical mode decomposition. Then the three-dimensional phase space representation (3D-PSR) is reconstructed for each IMF according to the time delay method. At last, the kernel radius of the corresponding 3D-PSR is defined, which aims to characterize the concentration degree of all the points in 3D-PSR. With the extracted feature KRF, we employ extreme learning machine and support vector machine as the classifiers to achieve the task of the automate epileptic seizure detection. Performances of the proposed method are finally verified on the Bonn EEG database.
KeywordsAutomatic seizure detection Electroencephalogram (EEG) Empirical mode decomposition (EMD) Phase space representation (PSR) Kernel-radius-based feature Extreme learning machine (ELM) Support vector machine (SVM)
This work was supported by the National Natural Science Foundation of China under Grant 61473223.
- 9.Niknazar, M., Mousavi, S.R.: A new dissimilarity index of EEG signals for epileptic seizure detection. In: Control and Signal Processing, pp. 1–5 (2010)Google Scholar
- 12.Ouyang, G., Li, X.L., Guan, X.P.: Use of fuzzy similarity index for epileptic seizure prediction. In: The 5th World Congress on Intelligent Control and Automation, Hang Zhou, China, vol. 6, pp. 5351–5355 (2004)Google Scholar
- 21.Zhu, G., Li, Y., Wen, P.: Epileptic seizure detection in eegs signals using a fast weighted horizontal visibility algorithm. Comput. Biol. Med. 115, 64–75 (2014)Google Scholar