Analýza a zpracování biometrických dat

Abstract

This thesis major focus is given to the area of environmental epilepsy sources detection, classification and prediction. Unimodal epilepsy detection is well known and described.

Multimodal detection for environmental source location on the other hand is not described in any study or literature. This research and thesis is motivated by practical experiences from the field of child neurology and people suffering from epilepsy. The primary issue is based unsatisfactory patient monitoring and patient’s environment monitoring. This environment often contains sources and triggers for reflex seizures.

This thesis presents solutions and algorithms for continuous mobile patient monitoring. Their brain activity is continuously monitored which classifies ictal and interictal brain activity. Environmental sources and stimuli's that can contain dangerous stimulants are monitored as well. Thanks to full system mobility the data can be recorded in the non-hospital environment as well.

Core of this thesis is GP-SVM method intended for ictal and interictal activity classification using combination of evolution strategies, genetic programming and machine learning algorithm SVM for the population member quality assessment. Amongst other original approaches is the idea of the continuous patient environmental reflex seizure provoking stimuli detection.