A new neural network model based on the LVQ algorithm for multi-class classification of arrhythmias

This paper describes the application of competitive neural networks with the LVQ algorithm for classification of electrocardiogram signals (ECG). For this study we used the MIT-BIH arrhythmia database with 15 classes. Three architectures were developed with a modular approach for classification. Compared with other methods that have been developed for classification of arrhythmias with this same database, the proposed approach produces very good results, because the entire database was used. Simulation results are presented, and a statistical test was performed to compare the three architectures which were very similar in the classification results.