I use maketopology script to generate topology file and the output is
#
# 3-state Bakis topology HMM with non-emitting last state
# These values are normalized so that rows sum to one.
#
# NO COMMENTS BETWEEN # OF STATES AND TRANSITION MATRIX
#
#
#Version number
0.1
# Number of states per model followed by transition matrix
4
3.0 1.0 0.0 0.0
0.0 3.0 1.0 0.0
0.0 0.0 3.0 1.0
# Last state has no outgoing arcs unless
# embedded in a sentence hmm structure
I read in the documentation that topology array contains booleans values to
indicate that states are connected or not , but in this matrix there is
numbers like 3
what is the meaning of that ?
and I have question about skipstate , when do I use it and why ?
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I read in the documentation that topology array contains booleans values to
indicate that states are connected or not , but in this matrix there is
numbers like 3 what is the meaning of that ?
I'm not sure what documentation are you talking about, but it seems wrong
and I have question about skipstate , when do I use it and why ?
Different HMM topologies are more suitable for different types of data. If
your observation sequencies typically contains jumps between non-sequential
states you might want to use skip state. It's rare situation for speech
though.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
3.0 is an initial rough approximation for the state transition probability.
The whole idea is that in transition matrix jump to next state probability is
bigger than stay in this state probability.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I use maketopology script to generate topology file and the output is
I read in the documentation that topology array contains booleans values to
indicate that states are connected or not , but in this matrix there is
numbers like 3
what is the meaning of that ?
and I have question about skipstate , when do I use it and why ?
I'm not sure what documentation are you talking about, but it seems wrong
Different HMM topologies are more suitable for different types of data. If
your observation sequencies typically contains jumps between non-sequential
states you might want to use skip state. It's rare situation for speech
though.
what is the meaning of number 3 here ?
3.0 is an initial rough approximation for the state transition probability.
The whole idea is that in transition matrix jump to next state probability is
bigger than stay in this state probability.
thanks