Factorization and particle swarm based sparse representation classifier for epilepsy classification implemented for wireless telemedicine applications

Abstract

Epilepsy is definitely a significant burden to the world because of its associated health related risks. For an epileptic patient to control and treat the unpredictable occurrence of seizures is quite a hectic task. A fast, efficient and versatile screening process is required that would aid the neurologists to diagnose and understand the treatment of the patient. Electroencephalography (EEG) is used traditionally for the in depth analysis of epilepsy. Since the recordings of the EEG are too long, processing such a huge data is difficult and therefore the dimensions of the data have to be reduced. In this paper, Variational Bayesian Matrix Factorization (VBMF) is employed to reduce the dimensions of the EEG data. The dimensionally reduced data is then transmitted through a Multiple Input Multiple Output Orthogonal Frequency Division Multiplexing (MIMO-OFDM) System. A Suitable Peak to Average Power Ratio (PAPR) Reduction scheme is engaged to reduce the Bit Error Rate (BER) and PAPR for the MIMO-OFDM System. At the receiver, the paper proposes a particle swarm based sparse representation classifier for the classification of epilepsy from EEG signals. The performance metrics analyzed here are specificity, sensitivity, time delay, quality values, performance index, accuracy, PAPR and BER.