On-line Detection of Patient Specific Neonatal Seizures using Support Vector Machines and Half-Wave Attribute Histograms

An efficient and effective Support Vector Machine for online seizures detection is presented. The kernel designed is based on features generated from bivariate histograms of EEG half-wave attributes. The training is on-line using a simple heuristic known as chunking. The case study presented illustrates the performance of the method on typical neonatal seizures.