, that is the partial probability of
being at state i at time t.Recognition System
of Japanese Characters
Forward Procedure
In this procedure a
forward variable is defined as , that is the partial probability of
being at state i at time t.
Forward Procedure
The figure above shows that state j at time t+1 can be reached by N different states. The algorithm is defined as:
1. Initialization
2. Recursion
3. Termination
Backward Procedure
In this procedure a
backward variable is defined as 
, that is the partial probability of
being at state i at time t.
Backward Procedure
The figure above shows that from state i at time t can reach N different states at time t+1.
The algorithm is:
1. Initialization
2. Recursion
Baum-Welch Algorithm
The paramet set are re-calculated as:
Last update 01.15.2004
These pages are maintained by Jorge Solis