HARDWARE IMPLEMENTATIONS OF MACHINE LEARNING TECHNIQUES FOR NEURAL SEIZURE DETECTION

Abstract

In this thesis an automatic seizure detection is proposed. For features extraction, more than 20 linear and nonlinear features are software implemented and tested to measure their efficiency in seizure detection. For classification block, two different algorithms are implemented: Artificial Neural Network (ANN) and Support Vector Machine (SVM). Support Vector Machine (SVM) training accelerators are also implemented using two different techniques: Gradient Ascent (GA) and Sequential Minimal Optimization (SMO). Finally, a new EEG dataset is extracted from rats in collaboration with a research team from the Faculty of Science, Cairo university and ONE lab.