Autoregressive estimation of closely-spaced spectra in the late potential region of the signal-averaged electrocardiogram as a function of segment width and ST slope

Absence of late potentials (LPs) in the terminal QRS and the early ST segment of the signal-averaged electrocardiogram has been found to be a reliable predictor of electrical stability in post-myocardial infarction (MI) patients. The short-time Fourier transform (STFT) has been used for the analysis of overlapping segments-of-interest within the SP region, the result being a time-frequency plot called a spectral-temporal map (STM). Three spectral estimation techniques based on autoregressive (AR) models are compared with the more conventional STFT method for STM analysis. Results indicate that AR techniques provide improved resolution of closely spaced spectra even for short analysis segments having sharp gradients