Identification of AR Time-series based on binary data

Romain Auber (1) , Mathieu Pouliquen (1) , Eric Pigeon (1) , Olivier Gehan (1) , Mohammed M'Saad (1) , Pierre Alexandre Chapon (2) , Sébastien Moussay (3)

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 study, the authors consider the identification of auto-regressive (AR) models for time-series from one-bit quantised observation sequences. The only available information is the fact that the samples of the time-series are lower or higher than a threshold of quantisation. This threshold may be different from zero. An identification algorithm is presented and analysed. A recursive formulation is proposed, an extension for the identification of a non-linear time-series is also proposed.

Keywords: Time series; quantisation (signal); Autoregressive processes; Nonlinear time-series; Quantisation; Identification algorithm; Autoregressive models; One-bit quantised observation sequences; AR time-series; Binary data

Domains: Engineering Sciences [physics]; Automatic

Romain Auber, Mathieu Pouliquen, Eric Pigeon, Olivier Gehan, Mohammed M'Saad, et al.. Identification of AR Time-series based on binary data. IET Signal Processing, 2019, 14 (1), pp.24-31. ⟨10.1049/iet-spr.2019.0152⟩. ⟨hal-02323954⟩