Activity Recognition from Binary Data

Romain Auber (1) , Mathieu Pouliquen (1) , Eric Pigeon (1) , Pierre Alexandre Chapon (2) , Sébastien Moussay (3, 2)

1 LAC - Laboratoire d'automatique de Caen
2 BodyCap Le compagnon santé connecté
3 INSERM - Institut National de la Santé et de la Recherche Médicale

In this paper, an algorithm for the activity recognition from binarized accelerometric data is presented. The particularity of the proposed algorithm is the use of binary data, this constraint on data is justified by the fact that using binary data allows to save battery and memory on the connected device. The objective of the present study is to show that it is possible to perform activity recognition from these binary data. The proposed algorithm uses Auto Regressive (AR) modeling and classification using Support Vector Machine (SVM). Some results on a real-data experiment is presented for the recognition of three activity.

Keywords: Activity recognition; Support vector machines; Accelerometers; Time series analysis; Data models; Training; Microsoft Windows; Autoregressive processes; Imagine recognition; Pattern classification; Binary data; Auto-Regressive model; Support Vector Machine

Domains: Engineering Sciences [physics]; Automatic

Romain Auber, Mathieu Pouliquen, Eric Pigeon, Pierre Alexandre Chapon, Sébastien Moussay. Activity Recognition from Binary Data. 2018 UKACC 12th International Conference on Control (CONTROL), Sep 2018, Sheffield, United Kingdom. pp.158-162, ⟨10.1109/CONTROL.2018.8516844⟩. ⟨hal-01827005⟩