HMM for isolated words recognition

HMM for isolated words data applying EM algorithm

In this project we would like to deal with training HMM for isolated words data applying EM algorithm. The testing phase is also considered using Viterbi algorithm. The results showed the performances which obtained by Matlab programming are similar to HTK's ones.
* Update for 2017-09-07
Before running these programs, please first prepare the training and testing data. Excerpts of TIDIGITS database can be obtained from this link:
http://www.ece.ucsb.edu/Faculty/Rabiner/ece259/speech%20recognition%20course.html
with the title of "isolated TI digits training files, 8 kHz sampled, endpointed: (isolated_digits_ti_train_endpt.zip)."
or, you may directly download the .zip file of training database from this link:
- training data: http://www.ece.ucsb.edu/Faculty/Rabiner/ece259/speech%20recognition%20course/databases/isolated_digits_ti_train_endpt.zip
- testing data:
http://www.ece.ucsb.edu/Faculty/Rabiner/ece259/speech%20recognition%20course/databases/isolated_digits_ti_test_endpt.zip
Please decompress all the data sets and locate them into the directory 'wav'.
We have just added some feature extracting functions that would help you to convert '.wav' files to '.mfc' (or '.mfcc') files (feature files)
Now you may run this project with only ONE click on the main function: 'EM_HMM_isolated_digit_main.m'