Personalized ECG monitoring:by sparse representation based feature extraction

Arrhythmia is a condition that may appear at any-time and anywhere. Unfortunately, the consequences are as well sometimes very serious for the suffering person. Therefore, the studies about automated ECG monitoring have gained a lot of attraction in the literature. This study focuses on the personalized monitoring aspect, which has been rarely investigated. In this thesis, we propose a novel sparse representation based feature extraction methods to be used with one class classification, which has potential to outperform the competing methods. To evaluate the performance, we run experiments with data from the MIT-BIH database and find that the proposed methods occasionally surpass the performance of the method that does not use the proposed feature extraction procedures and one class classification