D.S. Grigoryev, V.G. Spitsyn
The system for analysis and recognizing signals of electrocardiograms was developed. This system was tested on set of digital electrocar-diograms in different leads. The first part of system is two-level discrete wavelet–transform. This part decomposes signal functions of ECG on set of approximation and detail coefficients. Extracted approximation coefficients were used as inputs for next part of system. The se-cond part is an artificial neural network with one hidden layer. For training and testing neural network were used coefficients of signals from sets of 173 and 100 ECG respectively. The result of work of neural network is distribution of signals on given classes.
In this paper the objective is to recognize signals of normal sinus rhythm and signals with arrhythmia. Also, there were the researches of founding most optimal configuration of system. There was chosen type of wavelet function, level of wavelet transform, neural network training algorithm.