Multifractal characterization of synthetic ECG in the presence of coloured noise

The electrocardiogram (ECG) is a signal resulting from a complicated dynamical process. The paper describes an extension to a previous method of reconstruction and characterization of real ECG to a new "realistic" dynamical model of an ECG. A noise-free ECG is generated at 360 samples per second from the dynamical model, and contaminated with 10 dB of white, pink, or brown noise. The dynamical phase-space is then reconstructed. It is observed that the noise-free ECG has a Renyi fractal dimension spectrum (RS) bounded between 1 and 2. It is also observed that noise introduces a positive bias in the RS estimate. This bias is greater for uncorrelated noise, but more asymmetric with increasing correlation. These are the first characterizations of a realistic estimate of the noise-free ECG attractor. It is also the first analysis of the effect of noise correlation on the RS of ECGs in isolation of sampling effects. This demonstrates that a disagreement with the RS of a real ECG signal is not due to noise.