Classification of ECG signals using Hermite functions and MLP neural networks

Classification of heart arrhythmia is an important step in developing devices for monitoring the health of individuals. This paper proposes a three module system for classification of electrocardiogram (ECG) beats. These modules are: denoising module, feature extraction module and a classification module. In the first module the stationary wavelet transform (SWF) is used for noise reduction of the ECG signals. The feature extraction module extracts a balanced combination of the Hermit features and three timing interval feature. Then a number of multi-layer perceptron (MLP) neural networks with different number of layers and eight training algorithms are designed. Seven files from the MIT/BIH arrhythmia database are selected as test data and the performances of the networks, for speed of convergence and accuracy classifications, are evaluated.  Generally all of the proposed algorisms have good training time, however, the resilient back propagation (RP) algorithm illustrated the best overall training time among the different training algorithms. The Conjugate gradient back propagation (CGP) algorithm shows the best recognition accuracy about 98.02% using a little amount of features.