Next:
Reestimation of HMM Parameters
Up:
The standard HMM
Previous:
Using Gaussian Mixtures for
Contents
Forward/Backward Procedure
We wish to compute the probability of observation sequence
given the model
. The
forward procedure
for
is
Initialization:
(
13
.
9
)
Induction:
(
13
.
10
)
Termination:
(
13
.
11
)
The
backward procedure
is
Initialization:
(
13
.
12
)
Induction:
(
13
.
13
)