Fig. 2

Representative plots of the signal processing steps for deriving features of the iTUG, iSTS and FT tasks. In each subplot, vertical axis units correspond to the raw acceleration signal only. Other signals plotted have been rescaled for visualisation purposes. a The acceleration signal along the AP axis as recorded at the shank. Dashed and dotted lines represent the ACC signal processing steps for initial and final contact respectively, using the TK_GED algorithm [36]. b The acceleration signal recorded from the L5 sensor during the iTUG task. The start and end of the standing and sitting phases were identified as the local minima and maxima respectively of the double-differentiated filtered signal. c Events of the Sit-to-Stand cycle used to partition standing, sitting, complete stance and complete sit phases. ACC recorded from the RF was filtered and thresholds for standing and sitting were estimated (black dashed lines). d ACC recorded from the medial phalanx of the index finger during the FT task. Finger taps were identified as the local maxima of the ACC signal after filtering using the maximum-overlap discrete wavelet transform followed by the symmetric Teager-Kaiser Energy Operator