Multi-class ECG beat classification based on a Gaussian mixture model of Karhunen-Loève transform

摘要: Cardiovascular diseases are one of the main causes death around world. Automatic classification electrocardiogram (ECG) signals is paramount importance in unmanned detection a wide range heartbeat abnormalities. In this paper an effective multi-class beat classifier, based on statistical identification minimum-complexity model, presented. This methodology extracts from ECG signal multivariate relationships its natural modes, by means separation property Karhunen-Loeve transform (KLT). Then, it exploits optimized expectation maximization (EM) algorithm to find optimal parameters Gaussian mixture with focus being reducing number parameters. The resulting model thus estimation probability density function (PDF) that characterizes each type. Based above characterization was performed. experiments, conducted MIT-BIH arrhythmia database, demonstrated validity and, considering reduced size, excellent performance technique classify into different disease categories.