ECG classification based on sparse constrained nonnegative-matrix factorization and decision tree

In this paper, several data dimensionality reduction methods are compared. Then an ECG classification method is proposed, which employs the sparse decomposition of Nonnegative Matrix Factorization (SCNMF) for data dimensionality reduction, and Decision Tree for signal classification. The experimental results, in which five common heart diseases in the MIT-BIH database are used, indicate that the overall accuracy by the proposed ECG classification method reaches more than 99%. In addition, the employed data dimensionality reduction method can better retain the useful raw information and can save storage space.