A multi-resolution investigation for postural transition detection and quantification using a single wearable

•Evaluated wavelet algorithm for PT detection with a wearable on the lower back.•A 'one size fits all' algorithm for PT is insufficient.•Different PT types require different wavelet algorithm manipulations.•PT detection is accurate across a number of wavelets in controlled settings.•Temporal calculations unstable & require development prior to use in free-living.

BackgroundMulti-resolution analyses involving wavelets are commonly applied to data derived from accelerometer-based wearable technologies (wearables) to identify and quantify postural transitions (PTs). Previous studies fail to provide rationale to inform their choice of wavelet and scale approximation when utilising discrete wavelet transforms. This study examines varying combinations of those parameters to identify best practice recommendations for detecting and quantifying sit-to-stand (SiSt) and stand-to-sit (StSi) PTs.Methods39 young and 37 older participants completed three SiSt and StSi PTs on supported and unsupported chair types while wearing a single tri-axial accelerometer-based wearable on the lower back. Transition detection and duration were calculated through peak detection within the signal vector magnitude for a range of wavelets and scale approximations. A laboratory reference measure (2D video) was used for comparative analysis.ResultsDetection accuracy of wavelet and scale combinations for the transitions was excellent for both SiSt (87-97%) and StSi (82-86%) PT-types. The duration of PTs derived from the wearable showed considerable bias and poor agreement compared with the reference videos. No differences were observed between chair types and age groups respectively.ConclusionsImproved detection of PTs could be achieved through the incorporation of different wavelet and scale combinations for the assessment of specific PT types in clinical and free-living settings. An upper threshold of 5th scale approximations is advocated for improved detection of multiple PT-types. However, care should be taken estimating the duration of PTs using wearables.