Figure Captions
Figure 1: A) 0.2 fz sine wave with an amplitude of 100; B) Autocorrelogram of the 0.2 Hz sine wave illustrated in A. Data were sampled 2.0 Hz (lag is equivalent to sequential 500 msec sampling interval); C) Spectral analysis of the 0.2 Hz sine wave illustrated in A.
Figure 2: A) Chest circumference sampled 2.0 Hz; B) Autocorrelogram of data illustrated in A; C) Spectral analysis of data illustrated in A.
Figure 3: A) Heart period sampled 2.0 Hz from subject with prominent respiratory sinus arrhythmia; B) Autocorrelogram of data illustrated in A; C) Spectral analysis of data illustrated in A.
Figure 4: A) Heart period sampled 2.0 Hz from subject with little respiratory sinus arrhythmia; B) Autocorrelogram of data illustrated in A; C) Spectral analysis of data illustrated in A.
Figure 5: A) Sum of three pure sine waves of equal amplitude; B) Constituent periodicities of A; C) Autocorrelogram of data illustrated in A; D) Spectral decomposition of data illustrated in A.
Figure 6: A) Sum of three pure sine waves of unequal amplitude; B) Constituent periodicities of A; C) Autocorrelogram of data illustrated in A; D) Spectral decomposition of data illustrated in A.
Figure 7: A) Cross-correlogram of respiration and heart period data illustrated in Figure 8a; B) Cross-correlogram of data illustrated in Figure 8a transformed with the moving polynomial filter.
Figure 8: A) Simultaneously recorded heart period and respiration data; B) Spectral decomposition of data illustrated in A; C) Coherence spectrum of data illustrated in A.
Figure 9: A) Phase spectrum from the cross-spectral analysis of the respiration and heart period data illustrated in Figure 8a; B) Phase spectrum of the data illustrated in Figure 8a transformed with the moving polynomial filter.
Figure 10: A) Underlying rhythmic process to be sampled; B) Data generated by sampling the process illustrated in A too slowly (example of aliasing).
Figure 11: Spectral decomposition of white noise following a variety of filtering and detrending methods. A) no detrend; B) linear detrend; C) Successive-difference filter; D) Moving polynomial filter.
Figure 12: Moving polynomial procedure: Top panel is the unfiltered heart period series; Middle panel is the smoothed template series which is fit to the changing baseline and slow periodic processes; Bottom panel is the filtered heart period series illustrating respiratory sinus arrhythmia.
Figure 13: A) Heart period transformed by the moving polynomial filter superimposed on simultaneously recorded respiration; B) Spectral analyses of data illustrated in A.
Figure 14: A) Non-sinusoidal periodic component of signal illustrated in C; B) Sinusoidal component of signal illustrated in C; C) Complex periodic process.
Figure 15: A) Spectral decomposition of non-sinusoidal periodic signal illustrated in Figure 14A; B) Spectral decomposition of sine wave illustrated in Figure 14B; C) Spectral decomposition of complex periodic process illustrated in 14C;
Figure 16: Spectral decomposition of complex periodic process illustrated in Figure 14A following a variety of filtering and detrending methods. A) Linear detrending; B) Successive-difference filter; C) Moving polynomial filter.
Figure 17: Spectral decomposition of the heart period data illustrated in Figure 3A following a variety of filtering and detrending methods. A) No detrending; 8) Linear detrending; C) Successive-difference filter; D) Moving polynomial filter
Figure 18: A) Sine wave with amplitude of 50 msec.; B) Sine wave illustrated in Figure 18a superimposed on ascending trend; C) Sine wave illustrated in Figure 18a superimposed on descending trend.
Figure 19: Ascending trend with sine wave illustrated in Figure 18b detrended with different methods: A) Linear regression fit; B) Successive differences; C) Moving polynomial filter.