1: \begin{abstract}

2:   The Expectation Maximization (EM) algorithm is a versatile tool for model

3:   parameter estimation in latent data models. When processing large data sets

4:   or data stream however, EM becomes intractable since it requires the whole

5:   data set to be available at each iteration of the algorithm.  In this

6:   contribution, a new generic online EM algorithm for model parameter inference

7:   in general Hidden Markov Model is proposed. This new algorithm updates the

8:   parameter estimate after a block of observations is processed (online).  The

9:   convergence of this new algorithm is established, and the rate of convergence

10:   is studied showing the impact of the block-size sequence. An averaging procedure is

11:   also proposed to improve the rate of convergence.  Finally, practical

12:   illustrations are presented to highlight the performance of these algorithms in comparison to other online maximum likelihood procedures.

13: \end{abstract}

14: