Treatment of cardiac signal for a modeling by RBF

Abstract

In telemedicine, the transmission of the cardiac signal or for the diagnosis of an automatic Holter, it is important to model the heartbeat. Our aim in this work is the modeling of the ECG data by neural networks using Radial Base Function RBF. The treatment and cutting of ECG Holter helped us to find the best linear combination of five Gaussians that realizes this model. With a bank of Gaussian functions and using the algorithm Orthogonal Regressive Forward, we achieved an error of 10-4 in the initialization step. The optimization of this modeling is performed by the gradient algorithm.