Aim is to improve ECG signal de-noising method by dividing it in two stages, first stage involves QRS complex detection and second stage is filtering. In the first stage, this paper introduces MSMP and second stage filtering with AAF for ECG signal averaging.
Asynchronous Averaging and Filtering Method (AAF)
Signal averaging is a technique for combining the signals (images) recorded from various sources. Signal averaging is applied in varies applications to improve SNR of periodic signal mixed with random Gaussian noise. In cardiology, its main application is in the detection of ventricular late potential (Freedman, Gillis, Keren, Soderholm-Difatte, & Mason, n.d.).
Generally signal averaging technique is implemented wherein:
- 1.
Noisy ECGs are time aligned with the mean or median ECG signal
- 2.
Signal and noise are uncorrelated
- 3.
The nature of noise is random with a mean of zero is true.
Then averaging distorted signal m times improves SNR by a factor of m½ only when the random noise level of each portion of waveform has same characteristic and variability of waveform pattern is small (Herrera-Bendezu, Denys, & Reddy, n.d.) and (Alperin & Sadeh, 1986).
Signal averaging is based on the some characteristics of the signal and the noise like:
- 1.
Signal waveform should be periodic
- 2.
Noise must be random (not periodic) and uncorrelated with the signal
- 3.
Temporal position of each signal waveform must be accurately known (Valtino, Thompkins, & Nquyen, 1996).
AAF algorithm is basically a modified moving average filtering method. It includes two stages for two types of noise (EMG and baseline wander noise) cancellation. First M-point averaging for EMG noise cancellation and second double mean averaging for baseline wander cancellation.