Skip to main content
. Author manuscript; available in PMC: 2018 May 11.
Published in final edited form as: IEEE Trans Biomed Eng. 2016 Oct 25;64(8):1761–1771. doi: 10.1109/TBME.2016.2611862

Table II.

Feature Overview

Feature Name Abbr. Description
Entropy ENT A measure of the variability in the time that participants spend in the different locations recorded, defined as
ENT=i=1Npilogpi
where i = 1, …, N is the index of each location cluster extracted, N is the total number of clusters and pi is the percentage of time recorded cluster i.
Normalized Entropy NENT A variant of the ENT feature scaled to be in the range [0, 1], defined as
NENT=ENT/logN.
Location Variance LV An indication of how much the individual is moving between different locations based on the sum of statistical variances in the latitude and longitude, defined as
LV=log(σlat2+σlon2).
Home Stay HS The percentage of time that the participant is recorded in their home location.
Transition Time TT The percentage of all the time spent travelling between stationary locations in the data recorded.
Total Distance TD The sum of Euclidean distances between the consecutive location points recorded in the data, calculated as
TD=i=2N(xixi1)2+(yiyi1)2
where i = 1, …, N is the index of each of the total N pre-processed data points for the participant, and xi and yi are the distance east and north respectively from the assumed home location at coordinates (0, 0).
Number of Clusters NC The number of distinct location clusters extracted in the week-long data sections using the K -means method.
Diurnal Movement DM A measure of daily regularity quantified using the Lomb–Scargle periodogram to determine the power in frequencies with wavelengths around 24 h. The power spectral density (PSD) of the signal in selected frequencies with wavelengths between 23.5 h and 24.5 h is calculated and averaged as
E=i=1Npsd(fi)/N
where psd (f) is the PSD of the data at frequency f and fi for i = 1, …, N is the range of frequencies to use with wavelengths between 23.5 h and 24.5 h. This gives a measure of the energy in the spectrum with wavelengths around 24 h. The diurnal movement feature is defined as
DM=log(Elat+EIon)
where Elat and Elon are the energy values for the latitude and longitude of the recorded data respectively.
Diurnal Movement on Normalized Coordinates DMN Similar to the DM feature but calculated on a normalized set of coordinates, where the latitude and longitude are both scaled to have zero mean and unit variance within the period being classified.
Diurnal Movement on the Distance From Home DMD Similar to the DM and DMN features but calculated using the Euclidean distance from home, rather than latitude and longitude, normalized to have zero mean and unit variance within the period being classified.