2018年
Sequential Bayesian filters for estimating time series of wrapped and unwrapped angles with hyperparameter estimation
Journal of the Physical Society of Japan
- ,
- ,
- 巻
- 87
- 号
- 3
- 記述言語
- 英語
- 掲載種別
- 研究論文(学術雑誌)
- DOI
- 10.7566/JPSJ.87.034005
- 出版者・発行元
- Physical Society of Japan
The estimation of angular time series data is a widespread issue relating to various situations involving rotational motion and moving objects. There are two kinds of problem settings: the estimation of wrapped angles, which are principal values in a circular coordinate system (e.g., the direction of an object), and the estimation of unwrapped angles in an unbounded coordinate system such as for the positioning and tracking of moving objects measured by the signal-wave phase. Wrapped angles have been estimated in previous studies by sequential Bayesian filtering
however, the hyperparameters that are to be solved and that control the properties of the estimation model were given a priori. The present study establishes a procedure of hyperparameter estimation from the observation data of angles only, using the framework of Bayesian inference completely as the maximum likelihood estimation. Moreover, the filter model is modified to estimate the unwrapped angles. It is proved that without noise our model reduces to the existing algorithm of Itoh’s unwrapping transform. It is numerically confirmed that our model is an extension of unwrapping estimation from Itoh’s unwrapping transform to the case with noise.
however, the hyperparameters that are to be solved and that control the properties of the estimation model were given a priori. The present study establishes a procedure of hyperparameter estimation from the observation data of angles only, using the framework of Bayesian inference completely as the maximum likelihood estimation. Moreover, the filter model is modified to estimate the unwrapped angles. It is proved that without noise our model reduces to the existing algorithm of Itoh’s unwrapping transform. It is numerically confirmed that our model is an extension of unwrapping estimation from Itoh’s unwrapping transform to the case with noise.
- リンク情報
- ID情報
-
- DOI : 10.7566/JPSJ.87.034005
- ISSN : 1347-4073
- ISSN : 0031-9015
- SCOPUS ID : 85043241694
- Web of Science ID : WOS:000426732000009