A tool to extract meaningful health information from large accelerometer datasets. The software generates time-series and summary metrics useful for answering key questions such as how much time is spent in sleep, sedentary behaviour, or doing physical activity.


Dependancies include: unix, java 8 (Java 8 JDK) and python 3.7 (Anaconda’s Python 3 or installation via Brew should do the trick).

$ git clone git@github.com:activityMonitoring/biobankAccelerometerAnalysis.git
$ bash utilities/downloadDataModels.sh
$ pip3 install --user .
$ javac -cp java/JTransforms-3.1-with-dependencies.jar java/*.java

Getting started

To extract a summary of movement (average sample vector magnitude) and (non)wear time from raw Axivity .CWA (or gzipped .cwa.gz) accelerometer files:

$ python3 accProcess.py data/sample.cwa.gz
<output written to data/sample-outputSummary.json>
<time series output written to data/sample-timeSeries.csv.gz>

The main output JSON will look like:

    file-name: "sample.cwa.gz",
    file-startTime: "2014-05-07 13:29:50",
    file-endTime: "2014-05-13 09:49:50",
    acc-overall-avg(mg): 32.78149,
    wearTime-overall(days): 5.8,
    nonWearTime-overall(days): 0.04,
    quality-goodWearTime: 1

To visualise the time series and activity classification output:

python3 accPlot.py data/sample-timeSeries.csv.gz data/sample-plot.png
  <output plot written to data/sample-plot.png>

Output plot of overall activity and class predictions for each 30sec time window

The underlying modules can also be called in custom python scripts:

from accelerometer import summariseEpoch
summary = {}
epochData, labels = summariseEpoch.getActivitySummary("sample-epoch.csv.gz",
        "sample-nonWear.csv.gz", summary)
# <nonWear file written to "sample-nonWear.csv.gz" and dict "summary" updated
# with outcomes>

Citing our work

When describing or using the UK Biobank accelerometer dataset, or using this tool to extract overall activity from your accelerometer data, please cite [Doherty2017].

When using this tool to extract sleep duration and physical activity behaviours from your accelerometer data, please cite [Willetts2018] [Doherty2018] and [Walmsley2020].

[Doherty2017]Doherty A, Jackson D, Hammerla N, et al. (2017) Large scale population assessment of physical activity using wrist worn accelerometers: the UK Biobank study. PLOS ONE. 12(2):e0169649
[Willetts2018]Willetts M, Hollowell S, Aslett L, Holmes C, Doherty A. (2018) Statistical machine learning of sleep and physical activity phenotypes from sensor data in 96,220 UK Biobank participants. Scientific Reports. 8(1):7961
[Doherty2018]Doherty A, Smith-Bryne K, Ferreira T, et al. (2018) GWAS identifies 14 loci for device-measured physical activity and sleep duration. Nature Communications. 9(1):5257
[Walmsley2020]Walmsley R, Chan S, et al. (2020) Reallocating time from machine-learned sleep, sedentary behaviour or light physical activity to moderate-to-vigorous physical activity is associated with lower cardiovascular disease risk (preprint https://doi.org/10.1101/2020.11.10.20227769)


This project is released under a BSD 2-Clause Licence (see LICENCE file).

Indices and tables