Simultaneous positive sequential vectors modeling and unsupervised feature selection via continuous hidden Markov models

作者：

Highlights：

• A novel continuous hidden Markov model (HMM) is theoretically proposed by considering mixtures of generalized inverted Dirichlet distributions as its emission densities.

• We integrate an unsupervised localized features selection method into the proposed HMM in order to improve its performance for modeling high-dimensional data.

• A convergence-guaranteed algorithm based on variational Bayes is developed to learn the proposed model.

• The proposed continuous HMM is validated through both simulated data sets and a real-life application about human action recognition.

摘要

•A novel continuous hidden Markov model (HMM) is theoretically proposed by considering mixtures of generalized inverted Dirichlet distributions as its emission densities.•We integrate an unsupervised localized features selection method into the proposed HMM in order to improve its performance for modeling high-dimensional data.•A convergence-guaranteed algorithm based on variational Bayes is developed to learn the proposed model.•The proposed continuous HMM is validated through both simulated data sets and a real-life application about human action recognition.

论文关键词：Continuous hidden Markov models,Generalized inverted Dirichlet,Mixture models,Variational Bayes,Localized feature selection

论文评审过程：Received 25 February 2020, Revised 7 April 2021, Accepted 27 May 2021, Available online 15 June 2021, Version of Record 15 June 2021.

论文官网地址：https://doi.org/10.1016/j.patcog.2021.108073