Movement Prediction Using Accelerometers in a Human Population

Abstract

Summary We introduce statistical methods for predicting the types of human activity at sub-second resolution using triaxial accelerometry data. The major innovation is that we use labeled activity data from some subjects to predict the activity labels of other subjects. To achieve this, we normalize the data across subjects by matching the standing up and lying down portions of triaxial accelerometry data. This is necessary to account for differences between the variability in the position of the device relative to gravity, which are induced by body shape and size as well as by the ambiguous definition of device placement. We also normalize the data at the device level to ensure that the magnitude of the signal at rest is similar across devices. After normalization we use overlapping movelets (segments of triaxial accelerometry time series) extracted from some of the subjects to predict the movement type of the other subjects. The problem was motivated by and is applied to a laboratory study of 20 older participants who performed different activities while wearing accelerometers at the hip. Prediction results based on other people’s labeled dictionaries of activity performed almost as well as those obtained using their own labeled dictionaries. These findings indicate that prediction of activity types for data collected during natural activities of daily living may actually be possible.

Classification of human physical activity based on the raw accelerometry data via spherical coordinate transformation

“…The first set uses the aggregated signal to provide different measures of the energy expenditure, physical activity volume or its intensity (see: Bai et al (2016); van Hees et al (2013); ). The second set, which our work expands on, provides classification techniques for the human activity modes (see: Pober et al (2006); Mannini et al (2013); Krause et al (2003); Staudenmayer et al (2009) ;Trost et al (2012); Zhang et al (2012); Xiao et al (2015); Urbanek et al (2015); Straczkiewicz et al (2016)).…”

“…Zhang et al (2012) proposed an algorithm using combined methods that classified activity as walking, running, household, or sedentary activities. Xiao et al (2015) introduced technique based on an empirical basis called movelets designed specifically for the analysis of accelerometry data. The procedure classified 5 activity types: standing, lying, walking, upper body activities and getting up from a chair.…”

“…These approaches rely on the rotational independence of the activity counts and the vector length, which enables the signal comparisons for different devices and participants. A method which overcomes the aforementioned disadvantages is the movelets technique developed by Xiao et al (2015). This procedure detects short term activities such as walking a few steps around the house or getting up from a chair.…”

“…It is easy to notice that the comparison of both signals would be hard due to horizontal shifts and rescaling of the raw signal. In order to resolve this issue Xiao et al (2015) proposed a method of data normalization for different participants. However it demands additional information about participants standing and lying periods.…”

Stride variability measures derived from wrist- and hip-worn accelerometers

“…This shift is likely due to their ease of use, increased compliance of study participants, and improvements in size and battery life [1]. This shift raises new challenges to estimating gait parameters, as hands are involved in a much wider spectrum of activities, which results in higher complexity and increased within- and between-subject variability [23]. …”