Dissimilarity measure based on ordinal pattern for physiological signals

Complex physiologic signals may carry information of their underlying mechanisms. In this paper, we introduce a dissimilarity measure to capture the features of underlying dynamics from various types of physiologic signals based on rank order statistics of ordinal patterns. Simulated 1/f noise and white noise are used to evaluate the effect of data length, embedding dimension and time delay on this measure. We then apply this measure to different physiologic signals. The method can successfully characterize the unique underlying patterns of subjects at similar physiologic states. It can also serve as a good discriminative tool for the healthy young, healthy elderly, congestive heart failure, atrial fibrilation and white noise groups. Furthermore, when investigated into the details of underlying ordinal patterns for each group, it is found that the distributions of ordinal patterns varies significantly for healthy and pathologic states, as well as aging.