%0 Conference Proceedings %T Rank-based multiple change-point detection in multivariate time series %+ Laboratoire Information, Modèles, Apprentissage [Gif-sur-Yvette] (LIMA) %+ GIPSA - Communication Information and Complex Systems (GIPSA-CICS) %+ GIPSA - Signal et Automatique pour la surveillance, le diagnostic et la biomécanique (GIPSA-SAIGA) %A Harlé, Flore %A Chatelain, Florent %A Gouy-Pailler, Cedric %A Achard, Sophie %Z Conference Code:109162 %< avec comité de lecture %( IEEE Xplore %B EUSIPCO 2014 - 22th European Signal Processing Conference %C Lisbonne, Portugal %I European Signal Processing Conference, EUSIPCO %3 2014 22nd European Signal Processing Conference (EUSIPCO) %P 1337-1341 %8 2014-09-01 %D 2014 %K Time series %K Signal processing %K Signal detection %K Sampling %K MCMC method %K Gibbs sampling %K Bayesian networks %K Joint segmentation %K Bayesian inference %K Inference engines %K Hidden Markov models %K Stochastic models %K Rank Statistics %K Dependency structures %K artificial intelligence %K online learning %K Statistical tests %Z Mathematics [math]/Numerical Analysis [math.NA] %Z Computer Science [cs]/Signal and Image Processing %Z Statistics [stat]/Statistics Theory [stat.TH] %Z Statistics [stat]/Methodology [stat.ME]Conference papers %X In this paper, we propose a Bayesian approach for multivariate time series segmentation. A robust non-parametric test, based on rank statistics, is derived in a Bayesian framework to yield robust distribution-independent segmentations of piecewise constant multivariate time series for which mutual dependencies are unknown. By modelling rank-test p-values, a pseudo-likelihood is proposed to favour change-points detection for significant p-values. A vague prior is chosen for dependency structure between time series, and a MCMC method is applied to the resulting posterior distribution. The Gibbs sampling strategy makes the method computationally efficient. The algorithm is illustrated on simulated and real signals in two practical settings. It is demonstrated that change-points are robustly detected and localized, through implicit dependency structure learning or explicit structural prior introduction. %G English %2 https://cea.hal.science/cea-01844442/document %2 https://cea.hal.science/cea-01844442/file/article_FloreHarle.pdf %L cea-01844442 %U https://cea.hal.science/cea-01844442 %~ CEA %~ UGA %~ CNRS %~ UNIV-GRENOBLE1 %~ UNIV-PMF_GRENOBLE %~ UNIV-GRENOBLE3 %~ INPG %~ GIPSA %~ GIPSA-DIS %~ GIPSA-DA %~ INSMI %~ GIPSA-SAIGA %~ GIPSA-CICS %~ UGA-TEST-BIS %~ UGA-TEST-QUATER %~ DRT %~ TDS-MACS %~ LIST %~ DM2I %~ DIN