Time Series of Fuzzy Sets in Classification of Electrocardiographic Signals

A way of an application of time series of fuzzy sets to classification of QRS complexes of ECG signal for selected averaging of this signal is presented. After the formulation of the problem the notion of time series of fuzzy sets is recalled. The time series of fuzzy sets are created on the basis of the original noisy signal. The parameters of successive fuzzy sets are used as a feature vector for a classifier. In the presented paper, the l(2)-regularized iteratively reweighted least squares classifier and its kernel version are used. The MIT-BIH annotated ECG database is used in the experiments. The multi-fold cross-validation procedure using 100 pairs of learning and testing subsets are applied to validate the classification results. The obtained results (generalization error less than 1%) are very promising.