QRS detector pre-processing using the complex wavelet transform

Many approaches have been proposed to detect QRS complex in ECG signals. With wavelet transforms the algorithms proposed are based on discrete wavelet transforms (DWT), dyadic wavelet transforms (DyWT) and continuous wavelet transforms (CWT), all with real wavelet functions. This work presents a novel QRS detection CWT pre-processing scheme using a complex wavelet function. The method consists in calculating the absolute value of CWT of the 25 ECG signals using the complex Morlet wavelet function with dilatation levels from 1 up to 32. All the ECG signals used here were from MIT/BIH Arrhythmia Database. For comparison, is also applied to the ECG signals the pre-processing technique proposed in Pan and Tompkins (1985) and Hamilton and Tompkins (1986). Comparing all the graphics obtained in both methods for each ECG signal, it was chosen the CWT decomposition that has the greatest amplitude and also has the greatest analogy with the other pre-processing result. It can be concluded that the method proposed provided CWT decompositions with waveform very similar to those obtained by the other method. Also, the dilatation levels that gave the best are in the range of 18 and 27 (with an average value of 23 and std. deviation of 2.94).