Abstract

MENDOZA REYES, Miguel A; LORENZO GINORI, Juan V  and  TABOADA CRISPI, Alberto. Electrocardiographic signal classification using time-frequency representations. uct [online]. 2005, vol.9, n.35, pp.125-131. ISSN 1316-4821.

The analysis of electrocardiographic (ECG) signals constitutes an important element in the completion of different tasks for automated diagnosis, by means of computational algorithms. Classification of waveforms is an important objective within this analysis, for which various ECG processing algorithms in the time and frequency domains have been used. The non-stationary nature of the ECG signal has motivated also the application of techniques such as the time-frequency representations (TFR). The objectives of this work are to compare, different alternatives for the application of TFR to the classification of noisy QRS, and to introduce a classification method based in the use of complex values for a specific TFR, the Short Time Fourier Transform. The results of the application of this method are analyzed in comparison to a previous method introduced by the authors, also based in the use of TFR. The effectiveness of this method to classify noisy beats was determined by evaluating the use of various distance measures between beats contaminated with noise and reference patterns used for classification. A better performance was obtained during classification, for the noise types considered, when complex descriptors in the time-frequency domain were used.

Keywords : Beat Classification; Electrocardiographic Siguals; ECG; Time-Frequency Representations.