An efficient error-minimized random vector functional link network for epileptic seizure classification using VMD. (March 2020)

• Record Type: Journal Article
• Authors: Rout, Susanta Kumar; Biswal, Pradyut Kumar
• Is Part Of: Biomedical signal processing and control. Volume 57(2020)
• Publication Date: 2020-03
• Subjects: Electroencephalogram (EEG) signal -- Variational mode decomposition (VMD) -- Hilbert transform (HT) -- Error-minimized random vector functional link network (EMRVFLN) -- Digital implementation -- Epileptic seizure classification; 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.2019.101787
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: This paper presents an efficient algorithm for classification of Epileptic Seizure from normal, inter-ictal, and seizure EEG signals. A new classifier named EMRVFLN is proposed which is an improved version of both RVFLN and ELM. Efficient features are extracted after pre-processing from the EEG signal using VMD and HT. Also, this paper presents digital implementation of the proposed EMRVFLN classifier in FPGA environment. Two real time datasets i.e. Bonn university dataset and Neurology & Sleep Centre, Hauz Khas, New Delhi are used to validate the proposed method. Abstract: In this paper, variational mode decomposition (VMD), Hilbert transform (HT), and proposed error-minimized random vector functional link network (EMRVFLN) are integrated to detect and classify epileptic seizure from electroencephalogram (EEG) signals. VMD is applied to decompose the EEG signal into Band-limited intrinsic mode functions (BLIMFs). The five efficacious instantaneous features are computed using HT to construct the feature vector. Proposed EMRVFLN classifier is used to classify the epileptic seizure. The performances of the proposed EMRVFLN are compared with recently developed classifiers such as least-square support vector machine (LSSVM) and extreme learning machine (ELM). The combination of VMD and HT with proposed EMRVFLN classifier outperforms other state-of-the-art methods with classification accuracy of 100% for two class classification problem and 99.74% for three class classification problem. The remarkable classification accuracy facilitates the digital implementation of the proposed EMRVFLN classifier which may aid to design an embedded system for real-time disease diagnosis.