Tensor-based Uncorrelated Multilinear Discriminant Analysis for Epileptic Seizure Prediction

Epileptic seizure prediction explores the probability of forecasting the onset of epileptic seizure, which aids to timely treatment for patients. It provides a time lead compared to traditional seizure detection. In this paper, a spectral feature extraction is developed and the seizure prediction is performed based on uncorrelated multilinear discriminant analysis (UMLDA) and Support Vector Machine (SVM). To make best use of information in different dimension, we construct a three-order tensor in temporal, spectral and spatial domain by wavelet transform. And UMLDA implements the tensor-to-vector projection (TVP) with the minimum redundancy. The proposed solution employed 23 subjects' Electroencephalogram (EEG) data from Boston Children's Hospital-MIT scalp EEG dataset, each subject contains 40 minutes EEG signal. For the classification task of ictal state and preictal state, it exhibits an overall accuracy of 95%.