6
$\begingroup$

I know how to fit a hidden markov model to a data sequence, using the matlab-implementation of the baum-welch algorithm.

But what should I do if I do not have one data sequence, but a bunch of them? How can I fit to a set of sequences?

I think what I want to do is to maximize the likelihood of the HMM to output any of my sequences.

Is there a best practice for that kind of problem?

Improve this question
$\endgroup$
2

1 Answer 1

Reset to default
4
$\begingroup$

If you'd like to know the theory of doing this, it's covered in Rabiner's great paper "A tutorial to Hidden Markov models and selected applications in speech recognition" (Proc of the IEEE, 1989, 77(2), p.273; the full text available on multiple websites online - just google the name). As for whether there is an implementation in Matlab (or any other environment), I don't unfortunately know.

Improve this answer
$\endgroup$
2
  • $\begingroup$ Thank you, I had this paper on my desk and I should have read it way earlier... If you come across a good iterative algorithm to do it, let me know! $\endgroup$
    – Konstantin Schubert
    Mar 27, 2013 at 12:04
  • 1
    $\begingroup$ Found a matlab-package! Kevin Murphy's toolbox offers a possibility to fit to multiple samples of varying length and generally offers much more functionality than the functions available in the statistics toolbox. $\endgroup$
    – Konstantin Schubert
    Apr 27, 2013 at 23:17

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.