Automatically exported from code.google.com/p/indexinghelper
This was a project that three of us did at BYU to demonstrate how a hidden Markov model with multivariate Gaussian emission models could be used to transcribe handwriting in U.S. Census records.
Just remember this was a class project, though we did write up a paper ... though it looks like I need to find the final PDF to put in the repo. 🙂
And also be aware that my class mate who was supposed to write up the Viterbi algorithm (which is a fast dynamic programming algorithm) may have used a simpler brute force search instead. I have since rewritten the Viterbi algorithm for another project, but I don't think I incorporated it into this code. I will try to do so when necessary.