Classifying using the trained HMM parameters

The log-likelihood output of the software/hmm_reest.m program can be used as a classifier. If the number of iterations is specified as zero, a shortened version of the program is run, only running the forward procedure.
      [q, parm] = hmm_reest(parm, x, istart, nsamp, 0);
Since the program finds the total log likelihood for each record passed to it, the total likelihood will be the sum of the elements of q.



Baggenstoss 2017-05-19