Epliptic Seizure Detection and Classification Using Cumulative Sum Average Filter TT-Transform and Harmony Search Algorithm Based LLRBFN Model
Seizure detection becomes complex and difficult task for neurologists from electroencephalogram (EEG) signals. Therefore, It is essential to develop an automated detection and classification task to make detection and classification task easier for neurologist for the clinical diagnosis. This paper presents a novel hybrid Harmony search for optimization of weights of LLRBFNN (Local Linear Radial Basis Function Neural Network) model. The preprocessing has been adopted for noise removal and motion artifacts. Further, the Cumulative sum average filter and TT-transform has been used for noise removal and feature extraction from EEG elliptic seizure signals. Three features, namely power spectral density, Shannon entropy, and energy, were extracted. The dataset has been considered from University of Bonn database. The hybrid HS-LLRBFNN obtained an accuracy of 99.45% and the results are compared with HS-RBFNN, HS-LLWNN (Local Linear Wavelet neural network) models. Further, the results depicts the proposed model is appropriate for real-time seizure acknowledgement from EEG recording.