abstract
- The cardiovascular electrocardiogram signal presents noises that include low and high frequency components that interfere in the automatic identification and classification of the QRS peaks, P and T wave. Pre-processing techniques based on moving average and detrend function was used to smooth the ECG signal and remove local tendencies. Noise filtering methods based on FIR digital filters were implemented with the purpose of reducing interferences of the baseline and high frequency noise. For the localization of the R peak, an adaptive threshold, based on the Pan-Tompkins algorithm was applied. The other peaks, P-Q-S-T, were found from the search for highs and lows in a pre-determined zones. Pop-up windows were implemented to improve the accuracy of the semi-automatic localization. The developed algorithm obtained an accuracy of 98.09% in the QRS complex identification and 96.29% for the whole peaks under the MIT-BIH Arrhythmia Database.