A New Nonlinear Model Using Neural Networks For Generating Electrocardiogram Signals

The generation of electrocardiogram (ECG) signals by using a mathematical model has recently been investigated. One of the applications of a dynamical model which can artificially produces an ECG signal is the easy assessment of diagnostic ECG signal processing devices. In addition, the model may be also used in compression and telemedicine applications. It is also required that the model has capability to produce both normal and abnormal ECG signals. In this study, it is introduced a new method using radial basis function neural networks in a dynamical model based on McSharry model, to produce artificially the ECG signals. This new method has the advantage of capability to simulate a wider class of physiological signals (both normal and abnormal), compared to McSharry model. The simulation results are presented for normal ECG and three abnormal ones. The accuracy of the model has evaluated by using the error functions. The average of this error for a period of 100 seconds using 20 neurons is less than 2.5 percent for the four modeled cases (one normal and three abnormal).