Figure 3.

High-resolution (Canonical Time-series Characteristics 22 [Catch22]) wearable features from 3 different activity states. (A) Frequency polygons of the feature values based on the training set. The colors indicate activity states. (B) Pearson correlation coefficients between pairs of Catch22 features from different physical activity states (sleep, active, and sedentary). Two features from the active period (SC_FluctAnal_2_rsrangefit_50_1_logi_prop_r1 and SC_FluctAnal_2_dfa_50_1_2_logi_prop_r1) are uniformly 0; hence, correlation coefficients involving these 2 features are undefined (white squares).