Epileptic seizure detection using constrained singular spectrum analysis and 1D-local binary patterns

Early detection of epileptic seizures has a significant impact on patient outcomes. A novel pipeline for EEG-based epileptic seizure detection is here presented in which frequency factorisation is carried out on EEG signals by using constrained Singular Spectrum Analysis (SSA), coupled with one dimensional Local Binary Patterns (1-D LBP). The resulting frequency pattern transformation is classified via a Support Vector Machine (SVM) using Half Total Error Rate (HTER) in order to evaluate the performance of the proposed pipeline in a class-imbalanced context, with results compared against 1-D LBP on reference datasets. The results are tested against other comparable pipelines and demonstrate best-in-class performance.