next up previous contents index
Next: 8.7.1 Viterbi Training (HINIT) Up: 8 HMM Parameter Estimation Previous: 8.6 Single-Pass Retraining

8.7 Parameter Re-Estimation Formulae

 

  For reference purposes, this section lists the various formulae employed within the HTK parameter estimation tools. All are standard, however, the use of non-emitting states and multiple data streams leads to various special cases which are usually not covered fully in the literature.

The following notation is used in this section

 
		 N 		 number of states

S number of streams

tex2html_wrap_inline21622 number of mixture components in stream s

T number of observations

Q number of models in an embedded training sequence

tex2html_wrap_inline21630 number of states in the q'th model in a training sequence

tex2html_wrap_inline21634 a sequence of observations

tex2html_wrap_inline21636 the observation at time t, tex2html_wrap_inline21640

tex2html_wrap_inline21642 the observation vector for stream s at time t

tex2html_wrap_inline21648 the probability of a transition from state i to j

tex2html_wrap_inline21654 weight of mixture component m in state j stream s

tex2html_wrap_inline21662 vector of means for the mixture component m of state j stream s

tex2html_wrap_inline21670 covariance matrix for the mixture component m of state j stream s

tex2html_wrap_inline21678 the set of all parameters defining a HMM




next up previous contents index
Next: 8.7.1 Viterbi Training (HINIT) Up: 8 HMM Parameter Estimation Previous: 8.6 Single-Pass Retraining

ECRL HTK_V2.1: email [email protected]