Wrist-worn accelerometers (tri-axial signals over 9 full 24-hour days, 1GB/person) provide information on movement behaviours, such as activity and sedentary patterns as well as on sleep, covering a large range of components of circadian rhythm.
As algorithms to use accelerometer data were not available, their development has been an important element of the work undertaken, and these have been made available to the wider scientific community using the open access GGIR R package (http://cran.r-project.org/web/packages/GGIR/index.html).
We also invest in statistical approach to analyse these data comprehensively, using supervised or unsupervised machine learning to account for the large range of often correlated measures, and functional data analysis to account for the full distribution of activity intensity.
S Sabia, I Meneguel Danilevicz, J López García, S Vidil, A Dugravot
