Automated technique for EEG signal processing to detect seizure with optimized Variable Gaussian Filter and Fuzzy RBFELM classifier. (April 2022)

• Record Type: Journal Article
• Authors: Harishvijey, A; Benadict Raja, J
• Is Part Of: Biomedical signal processing and control. Volume 74(2022)
• Publication Date: 2022-04
• Subjects: EEG signals -- Signal classification -- Noise removal -- Filter algorithm -- Feature extraction -- Feature reduction -- Optimization algorithm; EEG Electroencephalogram -- VGF Variable Gaussian filter -- SSA Social spider algorithm -- RBF Radial basis function -- ELM Extreme learning machine -- K-PCA Kernel Principal Component analysis -- PSNR Power signal to noise ratio -- NN neural network -- SVM Support vector machine; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2021.103450
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: Epileptic seizure detection using automatic computerized method is presented. To obtain intensive smoothing of the signal, the Gaussian filter parameters are optimized and used for noise removal in EEG signals. Social spider optimization algorithm (SSA) is employed for optimizing the parameters of Gaussian filter. In feature extraction Empirical wavelet transform (EWT) is employed and the extracted features are reduced in size using kernel Principal component analysis (KPCA) technique. Feature reduction is performed to reduce computational time. In classification, to improve the classifier accuracy, Fuzzy Radial basis function (RBF) Extreme Learning Machine (ELM) is presented. Abstract: Epileptic seizure in patients is detected from EEG signals with the use of automatic signal classification techniques. The accurate detection of epilepsy is essential to reduce the risk of seizure related complications. However the available automatic signal detection techniques give poor sensitivity and accuracy. In this work, an automatic signal classification method for detecting seizure from EEG signal is presented for obtaining good classification results. The proposed work improves the performance of detection using Variable Gaussian filter (VGF) with social spider algorithm (SSA) (SSA-VGF), Empirical Wavelet Transform (EWT) feature extraction method, K- Principal component analysis (K-PCA) based feature reduction and Fuzzy logic embedded RBF kernel based ELM algorithm (FRBFELM). The SSA-VGF method is used for removing noise artifacts from the given EEG signals. EWT is employed for feature extraction and the size of extracted features is reduced using K-PCA method. Finally the signals are classified as normal signals and epileptic signals using FRBFELM classifier. The performance of the proposed method is evaluated by measuring the metrics; PSNR, accuracy, sensitivity, and specificity. The value of performance metrics obtained for the proposed work is 98.48%, 98.44% and 98.51%.