Raw and derived data for publication: Validity of accelerometry in step
detection and gait speed measurement in orthogeriatric patients

Authors: Keppler AM, Nuritdinow T, Mueller A, Hoefling H, Aigner G, Lederer C,
         Schieker M, Clay I, Daumer M, Böcker W,  Fürmetz J
DOI for paper: http://doi.org/10.1371/journal.pone.0221732
DOI for Data repository: http://doi.org/10.5281/zenodo.3153170

This is a description of the raw and derived data files generated in the
mentioned study.

* Raw data:

** raw.h5: raw acceleration data recorded at 100 Hz in HDF5 format

This file contains all acceleation raw data with 100 Hz fequency. The sensors
(when worn correctly) record x, y and z values corresponding to vertical,
lateral and back-forward acceleration. The 4th column (“KSS”) can be ignored, it
indicates if the belt buckle was a given time closed.  

To get acceleration gravity units the x, y or z value needs to be divided by
336. The KSS indicates whether the belt buckle was closed (1) or open (0). Note,
the order in which the data is returned depends on the programming language you
use, e.g. in R the dataset is returned as a matrix of 4 columns and n rows, in
phython it is n- columns and 4 rows. In the HDF5file itself, h5ls will list it
as a matrix with 4 rows and n columns.

The subject id is the group in the H5 file and the snesor name (either 'wheel'
or 'subject') the data set name. The 'wheel' contains the acceleratino of the
perambulator wheel that was used to derive reference ("gold standard") speed
averages per second. The 'subject' sensor contains acceleratino during walking.

** meta.cvs: sensor meta data
subject_id: subject id
subject_start: start time stamp (UTC) of the body-worn sensor
subject_end: end time stamp
wheel_start: start time stamp of the wheel-mounted sensor
wheel_end: end time stamp

** raw-wheel.csv:
subject_id: suject id for which this peramulator wheel records
interval_start: start of the interval (UTC)
interval_end: end of the interval (and interval should 1 second)
wheel_speed: average wheel speed for this interval

** raw-steps-annot.csv: unprocessed video step annotation
subject: subject id
actheel: index of heel strike in the body-worn sensor
acttoe: index oftoe-off in the body-worn sensor
utcheel: UTC time from video of heel strike
utctoe: UTC time from video of toe-off
side: left/right foot annotation

Note, from the index (acthell, acttoe) one can derive the UTC time: The meta.csv
file contains the start time per sensor, and each recording is 0.01 sec (100Hz).

** raw-parcours.csv: 
subject_id: subject id
test_label: lable of the parcour section
interval_start: video start of this section in UTC
interval_end: video end of this section

* Derived data:

** step-speed.csv: Steps and step speed for different algorithms and parcours
   sections:
subject_id: subject id
algo: algorithm used (stepslc, stepwave, wavesvr, video)
step_start: time stamp (UTC) for start of step 
duration: duratino of step in seconds
utcheel: video time stamp of heel strike (UTC), only for algo = 'video'
utctoe: video time stamp of toe-off (UTC), only for algo = 'video'
speed: step speed reported dby algorithm
GS_speed: "Gold Standard" speed (perambulator wheel speed)
freq: step frequency (in a window)
walk.ratio: walking ratio (in a window)
running: logical, is this a running step?
side: left/right
bout: bout id for this step (continuous un-interrupted walking bout)
test_label: the parcour section this step is in