A New Adaptive Wavelet Networks for ECG Recognition

Recognition of electrocardiogram (ECG) is an important area in intensive care. Automatic detection & classification of cardiac arrhythmias is important for diagnosis of cardiac abnormalities. Based on the wavelet transform theory, the wavelet networks have been wildly used for signal representation and classification. In this article, a new adaptive wavelet networks with one perceptron has been introduced for ECG signal recognition. An improved initialization approach was introduced and the relation between the number of the hidden layers and the astringency of the network also has been found. The network has been used for distinction between the normal beats and the premature ventricular contractions and has obtained high performance. In present work the ECG data are taken from MIT- BIH Arrhythmia database