![]() | Only 14 pages are availabe for public view |
Abstract In this thesis Artificial Neural Networks (ANN) are used to classify iectrocardiogram (ECG) signals. We depend on a database taken using a hand canner. The scanner output contains redundant information of tens or hundreds if kilos of bits. No network could deal with an input of that size so only ianificant features of the ECG signals are required. Two feature extraction methods are used to extract relevant features from ECG signals which are the Fast Cosine Transform (FCT) method and the Linear Prediction (LP) method. I n the FCT method, frames of only 600 samples were proven to be enough to be )resented to the neural network. In the LP method, we use a prediction order of Nvo, and the error signal is calculated which contains most of the features of the. ECG signals. We take a frame size of 200 samples and then we present 40 frames t o the network resulting in 8000 samples. The sequence of this thesis will be as follows: Chapter 1, is a general introduction concerning neural networks and ECG imals. |