0
$\begingroup$

First let me show this picture of my HMM Model.

enter image description here

Experimentally I get the High and Low values, whereas the states {|-1>, |0>, |1>} are unknown. Although I know that when I measure High signal I have |-1>. Also transition from |-1> to |1> are forbidden.

I heard that using a Markov Model here might be good to analyze a given time trace and get information about the hidden states. If I understand the Idea correctly the HMM gives me the most likely time trace given some transition probabilities. Then I optimize this time trace regarding my missing transition parameters ?

I'm glad for any help you can provide me. Also for good resources on how to implement Baum-Welch or Viterbi.

Okay thanks for the comment.

My question is:

What is the best and most time efficient way to get the transition rates between the hidden states ?

Improve this question
$\endgroup$
1
  • $\begingroup$ This is a good description of your situation / problem. Can you clarify what your question is exactly? $\endgroup$
    – gung - Reinstate Monica
    Commented Jan 26, 2017 at 13:56

1 Answer 1

Reset to default
1
$\begingroup$

here you go- [link] https://www.inf.ed.ac.uk/teaching/courses/asr/2012-13/asr03-hmmgmm-4up.pdf

I have understood HMM from these 3 videos- [link] https://www.youtube.com/watch?v=E3qrns5f3Fw
[link] https://www.youtube.com/watch?v=cjlhpaDXihE
[link] https://www.youtube.com/watch?v=5sGEF-e82yY

transition probabilities would be calculated in training phase.[Forward-backward Algo]

Improve this answer
$\endgroup$

Your Answer

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

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