DYNAMIC THRESHOLDING CONSTANT COMPUTATION FOR DETECTION OF QRS-COMPLEXES IN ELECTRO-CARDIO GRAM SIGNALS USING GENETIC ALGORITHM AND VARIABLE WAVE APPROXIMATION

Abstract

Detecting QRS-complexes from ECG waveform accurately, can now be treated as a classical problem. Many of the traditional QRS-detection algorithms somehow uses an empirically derived, universal thresholding constant to separate out QRS-complexes from Non-QRS region, after some processing for different patient records. For instance the processing of an ECG signal may comprises of noise-filtering the raw signal, differentiating, squaring, running a moving window integral, normalizing and finally using a constant threshold for discrimination. The method of thresholding works best after normalization. Unfortunately, in cases, where a noise peak is of greater amplitude (than QRS-complexes) the normalization process scales-up the noise signal to unity and suppresses the QRS-complexes (may be, lower than the thresholding constant). This may even result in neglecting the actual QRS-complexes and counting the noise peak as the only QRS-complex in the signal. To avoid this problem if the thresholding constant is set to a lower level some T-waves may accidentally be detected as QRS-complexes.