Abstract
Accurate detection of anomalies within electrocardiogram signals is a problem with significance in the field of medicine. An accurate solution to this problem that is both cost and time effective would have applications in cardiovascular medicine world-wide. This paper presents a two-part method for electrocardiogram anomaly detection. The first step in the proposed method is to de-noise the chosen signal through the process of wavelet decomposition and reconstruction. The second involves inputting the de-noised signals into a feed-forward artificial neural network. The network will be trained on two types of signals: a set of signals that are known to contain anomalies and a set of healthy control signals. Using these two sets of signals, the network will be trained to identify set-specific patterns in each of the two signal sets. The objective presented is to train the network to be able to accurately differentiate between healthy signals and signals containing anomalies.