Epileptic Seizure Detection and Classification based on RQA Features

V. Lakshmana Rao, Dr.K.V. Ramana and Dr.P. Krishna Subba Rao

To examine the dynamical systems, a non linear method of data analysis called Recurrence quantification analysis (RQA) was observed. It evaluates the amount and span of recurrences of a dynamical framework introduced by its phase space trajectory. The recurrence evaluation analysis was created so as to measure distinctively showing recurrence plots (RPs) in the view of the little scope structures that are involved. Recurrence plots are mechanisms which envision the repeated conduct of the phase space trajectory of dynamical systems. Generally, these plots contain single dots and lines which are corresponding to the mean diagonal or which are vertical/horizontal. Lines corresponding to the LOI are alluded to as diagonal lines and the vertical structures as vertical lines. The lines compare to the regular behaviour of the stage space trajectory: while the diagonal lines depicts such segments of the stage space trajectory which run equal for quite a while, the vertical lines depicts to portions which remain in the same phase space region for quite a while. RQA on EEG accounts and their sub groups: delta, theta, alpha, beta, and gamma for epileptic seizure detection. RQA was acquired since it doesn't require suspicions about stationarity, length of signal, and commotion. In this we prepared the EEG signal in both time and recurrence spaces and applied a Chebyschev filter for preprocessing the signal, at that point, by utilizing DWT to decay of the intial EEG into its five constituent sub groups enables better distinguishing proof for the dynamical system of EEG signal. This prompts better categorization of the database into three groups: Healthy disciplines, epileptic disciplines during a no-seizure interval (Interictal) and epileptic subjects during a seizure course (Ictal). In our work we will Analyze EEG signal by RQA techniques, which comprise of the procedure of Wavelet decomposition, Reconstruction and analysis. At that point the fature values are grouped utilizing ECOC classifier, which is mode of nearest neighbor Classification procedure.

Volume 12 | 05-Special Issue

Pages: 251-261

DOI: 10.5373/JARDCS/V12SP5/20201755