Abstract
Power spectrum analysis of heart rate variability was calculated using seven methods: (1) periodogram; (2) Levinson-Durbin algorithm; (3) Burg algorithm; (4) Marple least-squares algorithm; (5) Pisarenko harmonic decomposition; (6) extended Prony methods, and (7) Prony spectral-line decomposition. Heart rate was monitored for several hours of Holter recordings in control conditions and under induced changes in the autonomic nervous system. The least satisfactory results were obtained by the periodogram method using a direct FFT. Decomposition methods have not represented well the changes in heart rate, especially when periodicities in the rate were more than three distinct frequencies but contaminated with wideband noise. Levinson-Durbin, Burg and Marple algorithms gave similar power spectra of which the 0.05-0.1 Hz frequency band gave a good representation of the activity of the autonomic nervous system.
Original language | English |
---|---|
Pages (from-to) | 401-404 |
Number of pages | 4 |
Journal | Computers in Cardiology |
State | Published - 1987 |
Externally published | Yes |