MethodPDF Available

Activity Intensity Classification Method based on Tri-Axial Accelerometry Data

Authors:
  • South-Eastern Finland University of Applied Sciences

Abstract

This white paper provides a quick overview of a method related to the activity intensity classification method based on tri-axial accelerometry data utilised in the Fibion device.
© 2019 Fibion Inc.
All rights reserved
www.fibion.com
[1] Yang Y, Schumann M, Le S, Cheng S. 2018. Reliability and validity of a new accelerometer-based device for detecting physical
activities and energy expenditure. PeerJ 6:e5775 https://doi.org/10.7717/peerj.5775
Activity Intensity Classification
Method based on Tri-Axial
Accelerometry Data
This white paper provides a quick overview of
a method related to the activity intensity
classification method based on tri-axial
accelerometry data utilised in the Fibion
device.
Commonly, physical activity is divided into
sedentary behaviour (SB), light-intensity physical
activity (LPA), moderate-intensity (MPA), and
vigorous-intensity activity (VPA). Sedentary
behaviour has earlier been classified as all activities
that have energy expenditure less than 1.5 metabolic
equivalents (METs). Nowadays it has been more
accurately and relevantly classified based on
posture, in a way that SB is laying down and sitting.
Standing still is considered as the lowest intensity
physical activity.
Physical activities are then classified into different
categories based on the intensity of the activity (or
more literally the energy expenditure of the
activity). Activity is classified based on multiples of
resting metabolic rate in a way that between all
activities under 3 METs (excluding sedentary
behaviours) are classified as light-intensity physical
activity (LPA), as moderate-intensity (MPA)
between 3 and 6 METs, and as vigorous-intensity
activity (VPA) over 6 METs. Furthermore,
moderate- and vigorous-intensity activity is often
combined to form category moderate-to-vigorous
intensity activity (MVPA). So basically, this
classification method provides intensity classes that
are relative to resting energy expenditure but not to
maximal capacity of the individual.
For a good quality physical activity research, it is
important that measurement method is capable of
distinguishing accurately between different
intensity categories of physical activity as each of
these intensity categories causes a different kind of
stimulus for the body, therefore, providing
differential health benefits or risk reductions against
different adverse health conditions.
Normally, accelerometry-based physical activity
intensity classification is based on the magnitude of
accelerations. It is based on a generic notion that the
higher the accelerations, the more intensive the
physical activity. This approach has considerable
limitations, especially in activities, in which the
device movement (or lack of movement) is
disproportional to the intensity of the activity. While
the magnitude of acceleration may be related to
activity intensity rather linearly within a certain
activity type (e.g. walking), the same is not true
between different activity types. For example,
energy expenditure (i.e. intensity) of standing is
higher than that of sitting even if the magnitudes of
accelerations are the same as in upright position
since antigravitational muscles need to work to
maintain the position. Furthermore, for example,
during cycling, the accelerations may be smaller
than during walking, even when energy expenditure
of cycling is higher than in walking, as cycling does
not produce ground reaction accelerations of similar
magnitude as walking.
The described method does works on different
principles and does not have the disadvantages
described above as it takes into consideration the
differing energy expenditure during different
activity types. The method utilises advanced
algorithms that take into account different postures,
activity types, their intensity as well as the
anatomical and physiological differences between
individuals. Despite the high accuracy and
complicated algorithms that run in the background,
the method is very straightforward and simple to use
in various settings ranging from laboratory to free-
living.
Figure 1 shows a schematic overall description of
the activity intensity classification method. In the
method, tri-axial acceleration data is first analysed
to get the variables related to the orientation of the
device, acceleration in three different axes,
direction, as well as the magnitude of the resultant
acceleration. Furthermore, the change of orientation
and the frequency/cadence of accelerations are
© 2019 Fibion Inc.
All rights reserved
www.fibion.com
[1] Yang Y, Schumann M, Le S, Cheng S. 2018. Reliability and validity of a new accelerometer-based device for detecting physical
activities and energy expenditure. PeerJ 6:e5775 https://doi.org/10.7717/peerj.5775
analysed. This data is then used to classify activities
to different activity types as the first stage of activity
classification. In the next stage, characteristics,
related to each activity type, that affects EE are then
detected and analysed. These analyses are
performed by the firmware in the device on the fly
and stored into the memory of the device.
After the measurement period, data from the device
is uploaded to the Fibion cloud and the background
information of the participant is provided in the data
upload form. Then activity type classification and
intensity detection are being refined as cloud
processing algorithms are analysing the data as a
whole’ and taking into account also the background
information of the participant. Background
information of the participant (age, sex, height,
weight) is used to calculate estimations of
anatomical and physiological characteristics that
affect energy expenditure of an individual (e.g. body
mass index, muscle mass, estimated resting
metabolic rate). In the next stage, EE of each
activity is being calculated with activity type-
specific algorithms that take into account the
intensity of activity, background information, and
calculated anatomical and physiological
characteristics.
Then activities are divided broadly to sedentary
behaviours and physical activity based on the
posture in a way that all upright positions and
cycling are considered as physical activity and
laying down and sitting are classified as sedentary
behaviours. In the next stage physical activity is
divided into light-, moderate- and vigorous-
intensity activity. In this classification, metabolic
equivalent values are observed on a second-by-
second basis and each time point of physical activity
is categorised into one of the three intensity
categories. Moderate- and vigorous intensity
activity is combined to provide variable for MVPA.
The activity intensity classification has been
validated in a well-designed scientific research
setting. The results show that the model provides
accurate estimations of physical activity intensity
categories during challenging and varying tasks of
everyday life [1].
Figure 1. A simplified illustration of the analysis procedure.
ResearchGate has not been able to resolve any citations for this publication.
ResearchGate has not been able to resolve any references for this publication.