Spectral Analysis of Heart Rate Without Resampling

GB Moody
Harvard-M.I.T. Division of Health Sciences and Technology, Cambridge, MA, USA

This article originally appeared in Computers in Cardiology 1993, vol. 20, pp. 715-718 (IEEE Computer Society Press). Please cite this publication when referencing this material. Also available are PostScript and PDF versions of this article, a portable C implementation of the spectral analysis algorithm described below, and a brief user's guide. The algorithm for deriving instantaneous heart rate, as provided in the article (following the bibliography) is a simplified version of ihr, for which a brief user's guide is available here.

Abstract:

Standard methods of estimating the power spectral density (PSD) of irregularly sampled signals such as instantaneous heart rate (HR) require resampling at uniform intervals and replacement of unusable samples. The Lomb periodogram is a means of obtaining PSD estimates directly from irregularly sampled time series, avoiding these requirements. This paper compares Fourier, autoregressive, and Lomb PSD estimates from synthetic, real, and noise-corrupted real heart rate time series, and examines systematic differences among these estimates. An algorithm is presented for obtaining a heart rate time series suitable for Lomb PSD estimation from an RR interval time series with included ectopic beats and erroneous measurements. The paper concludes with a brief survey of other applications of the technique, such as estimation of respiratory frequency from a time series of beat-by-beat measurements of the mean electrical axis.