Identification of local model network application to electrocardiogram segmentation
Résumé
In this paper, a method for segmentation and modelling of an ECG signal using a local model network is proposed. The purpose is to perform a diagnosis of these signals in order to detect a particular pathology. From a technical point of view, the main problems to solve are the parameter estimation of local models, the parameter estimation of the commutation function, the choice of the number of local models and the structure of each local model. In this paper, the two first problems are addressed and solutions are proposed.