XPost: comp.ai.nat-lang   
   From: miltonsarria@gmail.com   
      
   I am working in the implementation of an algorithm for the inclusion of   
   explicit duration density in hidden markov models for automatic speech   
   recognition,   
   in the case where the observation sequences are discrete, i'm basing   
   the work in the article   
   [1] "A tutorial on hidden markov models and selected aplications in   
   speech recognition" of Lawrence R. Rabiner (pages 13 (269) and 14   
   (270)),   
   and   
   [2] "continuously variable duration hidden markov models for automatic   
   speech recognition" of S. E. Levinson.   
   The problem is that the "forward" and "backward" algorithms that are   
   described in [1] are different  to the ones described in [2], as [2] is   
   cited in [1], I am searching the reestimation formulas consistent with   
   the notation in [2] but I can't find the reestimation formula for the   
   observation symbol probability distribution in estate j (B={bj(k)}),the   
   initial  probability distribution and the duration probability   
   distribution in state j (Pj(d)), since I am working in the discrete   
   case, and [2] is about continuous Hmm's.   
   A article that I have not found is  "Explicit Modelling of State   
   Occupancy in Hidden Markov Models for Automatic Speech Recognition" of   
   Russell, M. J. and Moore, R. K., if somebody has it please send it to   
   me, THANKS.   
      
   [ comp.ai is moderated.  To submit, just post and be patient, or if ]   
   [ that fails mail your article to , and ]   
   [ ask your news administrator to fix the problems with your system. ]   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)