PCA based selective mapping technique for reduced PAPR implemented for distributed wireless patient monitoring epilepsy classification system

Abstract

To assess the neurological conditions, non- invasive techniques for recording activities of the brain are used widely. The recording of the brain activity helps us to understand the brain functions properly. The scalp Electroencephalography (EEG) recordings act as a fundamental tool to trace the pathological brain activity and so it is used widely to treat the patients affected with epilepsy. Since the recordings of the EEG are quite long and hectic to process, Linear Graph Embedding (LGE) is used to reduce the dimensions of the EEG data in this paper. It is then transmitted to the Space Time Trellis Coded Multiple Input Multiple Output Orthogonal Frequency Division Multiplexing (STTC MIMO OFDM) System. As the system suffers from a high PAPR, Principal Component Analysis Based Selective Mapping Technique (PCA-SLM) is used to reduce the PAPR. At the receiver the Gaussian Kernel Based Support Vector Machine (GK-SVM) is employed to act as a post classifier to classify the epilepsy risk levels from EEG signals. Along with the classification results, the Bit Error Rate (BER) is also analyzed in the receiver. The performance metrics here are Specificity, Sensitivity, Time Delay, Quality Values, Accuracy and Performance Index.