Fig 1.
Illustration of coarse-graining procedure using mean of data points in a non-overlapping window equal in length to the time scale.
Fig 2.
(a) Original times series (b) Coarse-grained time series (c) Data symbolization process (d) Histogram of symbol sequences.
Fig 3.
Mean ± SD values of the (a) MSE and (b) MNCSE computed from40 different WGN and 1/f noise signals.
Fig 4.
MSE (a-f) and MNCSE (g-l) as a function of data length, computed from 40 realizations of WGN and 1/f noise signals.
Table 1.
Mean ranks, corresponding p-values and area under ROC curve for comparison of MNCSE and MSE at temporal scales 1 to 15 for quantifying the dynamics of NSR, CHF and young and elderly subjects.
Fig 5.
Mean ranks for comparison of a) MNCSE and b) MSE at temporal scales 1 to 15 for quantifying the dynamics of pathological with disease severity.
Fig 6.
Mean ranks for comparison of a) MNCSE and b) MSE at temporal scales 1 to 15 for quantifying the dynamics of pathological with disease severity.
Fig 7.
Classification accuracy (CA), computed using (a)10 by 10 FCV for separating NSR and CHF subjects (b) 10 by 10 FCV for separating NSR young and NSR elderly subjects (c) leave-one-out cross validation for separating NSR and CHF subjects (d) leave-one-out cross validation for separating NSR young and NSR elderly subjects.
Fig 8.
(a) RR interval time series from healthy subject (b) Time series obtained by excluding artifacts greater than 2s. (c) MNCSE analysis (d) MSE analysis.