An improved weighted signal averaging method for high-resolution ECG signals

Proposes a new technique to estimate the coefficients for the weighted averaging of high-resolution ECG (HRECG) records. The purpose was to minimize the number of beats to average in records contaminated with non-stationary noise. The following averaging methods were studied: (1) linear averaging of all beats (LA), (2) linear averaging rejecting very noisy beats (RA), and (3) weighted averaging (WA). For WA, two techniques for noise variance estimation in each beat were analysed: (a) a noise estimation in a 60-ms window on the ST segment (WA-W), and (b) a new algorithm that estimates the noise variance from the difference signal obtained by subtracting the averaged beat from the individual beat (WA-D). All methods were tested in simulated HRECG records contaminated with several types of noise. We conclude that WA methods minimize the number of averaged beats compared with the LA and RA methods in situations of non-stationary noise. However, the proposed WA-D method shows a higher effectiveness than the WA-W method due to a better estimation of the noise variance in individual beats