The measurement of heart rate variability has become an important indicator in the diagnosis of certain diseases due to the strong relationship between the nervous system and the cardiovascular system. However, this measurement may be biased by the presence of ectopic beats, which do not originate in the sinoatrial node, causing alterations in the heart rate. The paper presents a set of techniques to address the nonlinear nature of the ECG signals and the strong influence that may cause noise in their processing. Among the techniques used is the Poincaré plot which identifies changes in the RR intervals which may possibly be associated with ectopic beats. In the feature extraction process, a third-order cummulant function was used, which provides noise immunity and allows the extraction of certain characteristics of each type of ectopic beats. Finally, the classification stage is performed using neural networks, which allow for an adequate solution to the problem of classification of ectopic beats.