Ph.D. Tezi Görüntüleme

Student: Yusuf SEVİM
Supervisor: Asst. Prof. Dr. Ayten ATASOY
Department: Elektrik-Elektronik Müh.
Institution: Graduate School of Natural and Applied Sciences
University: Karadeniz Technical University, Turkey
Title of the Thesis: Fetal Electrocardiogram and Maternal Electrocardiogram Signals Separation Using Approach to Pseudo-Additivity
Level: Ph.D.
Acceptance Date: 12/8/2009
Number of Pages: 145
Registration Number: Di708
Summary:

      Electrocardiogram signals can be generated by the heart’s electrical actions and can be detected using electrodes over skin. These biological signals are amplified by using amplifiers

and can be processed with analog or digital signal processing techniques. These processed signals serve as an excellent tool for heart diseases analysis. Most important property of using skin electrodes is that it is a non-invasive technique, meaning that this signal can be measured without entering the body at all. Despite this facility, non-invasive methods include problems such as mixing with other physiological signals and noise.

      Fetal electrocardiogram visualizes the electrical physiological activity of the fetal heart and it contains important indications about the health and condition of the fetus. Obtaining of the fetal electrocardiogram can be carried out through skin electrode like adults. But the same problems appear evidently during the obtaining of fetal electrocardiogram. In addition these problems, mother’s electrocardiogram signal can be also problem for fetal electrocardiogram signal since its amplitude is much higher than that of the fetus. It is impossible to recover the wanted fetal electrocardiogram components from the mixed potential recordings by conventional filtering techniques.

In this thesis, the purpose is to determine the appropriate algorithms for fetal electrocardiogram extraction by means of blind source separation algorithms and find a new blind source separation algorithm. As the first step, artificially generated electrocardiogram signals were mixed and used to find suitable algorithms. Three main algorithms of blind source separation; FastICA, JADE, non-parametric ICA, and new non-parametric ICA algorithms compared for this purpose.

      The results obtained indicate that non-parametric algorithms can be employed to preprocess electrocardiogram signals to separate the effects of maternal electrocardiogram signal and interference signals.

      Key Words: Independent Component Analysis, Blind Source Separation, Electrocardiogram, Entropy, Tsallis Entropy, Shannon Entropi, Information Theory, Relative Entropy