Axivity AX3 Light and Temperature Extractor

A tool created to extract the raw light and temperature data from Axivity AX3 wrist-worn accelerometers (as used in the UK Biobank). This was created whilst working at the Complex Trait Genomics team at the University of Exeter Medical School (http://www.t2diabetesgenes.org/who-we-are/). Please contact James Christie (jlcc201@exeter.ac.uk) or Andy Wood (a.r.wood@exeter.ac.uk) for more information.

Features:

• convert large .cwa binary files to .csv
• light data extraction
• temperature data extraction
• central moving average of data (variable window size)
• summarise individuals light or temperature data (useful for summarising many individuals in to one file)
• framework exists to extract other data such as raw accelerometer readings

Install

1. clone this repo
   git clone https://github.com/jlc-christie/axivity-ax3-tool.git
2. compile with prefered c++11 compatible compiler
   e.g. g++ ax3.cpp -o ax3 -O3, if using g++

Usage

./ax3 (-l|-t)  -i <input file> -o <output file> [-a <window size>] [-s <summary file>]
    -l -- Light mode
    -t -- Temperature mode
    -i -- Path to .cwa input file
    -o -- path to .csv output file to be generated
    -a -- Use Central Moving Average on data read
    -s -- Calculate summary statistics

Example:

Extract the light (-l) data from the input file my_raw_data_file.cwa, use a central moving average (-a) of 5 minutes (300 seconds) and save the output to light_data.csv. Also, append summary statistics about the light data to the file summary_statistics.csv.

./ax3 -l -i my_raw_data_file.cwa -o light_data.csv -a 300 -s summary_statistics.csv

Plotting

This repo also contains a small gnuplot script to either interactively plot the results or save the plots to files. Use -e "filename='out.csv'" to pass in the data file, making sure that the -e flag comes before the script file. Append -p to display the interactive plot and finally pass another argument outfile if you want the plots to be saved to a file (recommended, as interactive plot is relatively slow). Examples:

1. Display interactive plot, don't save plot images
   gnuplot -e "filename='out.csv'" plot.gpl -p
2. Display interactive plot and save plot images
   gnuplot -e "filename='out.csv';outfile='my_plot'" plot.gpl -p
   creates 2 files, my_plot.png and my_plot.ps, ps can be converted easily to pdf if required (e.g. ps2pdf my_plot.ps my_plot.pdf)
3. Don't display interactive plot, just save images to files
   gnuplot -e "filename='out.csv';outfile='my_plot'" plot.gpl

Example output from plot:

Summary Statistics

As mentioned above, the -s flag followed by a summary file filename, will calculate the mean and standard deviation of light or temperature (depending on which mode it is in) for the whole day as well as hourly. Because this functionality is meant to be used as part of a batch script which appends to the same file, there is no header. The data is comma seperated and the format is as follows:

filename	mean	iqr	std_dev	hour_0_mean	hour_0_std_dev	hour_1_mean	hour_1_std_dev	...	hour_23_mean	hour_23_std_dev
27492749	20.558	120.000	1.381	21.550	1.656	21.229	1.524	...	20.972	1.120