A hidden Markov Model (hmm) based speaker identification system using mobile phone database of North Atlantic Treaty Organization (nato) words

Speaker Recognition System Architecture:
The architecture for developing the speaker recognition
system is shown in figure 1. It consists of two modules, training module and testing module [4].Training
module generates the system model which is to be used during testing. Training module generates the system 
model which is to be used during testing. The various procedure used during the automatic speaker
recognition process [5] are shown in fig.1
Training Phase:
speaker
Testing Phase:
Test
Speaker
Recording
NATO words 
from different 
speakers
Feature
Extraction
(MFCC)
Database of 
features for each 
speaker
HMM
Model
Recording of
NATO words
Feature Extraction 
(MFCC)
HMM Model
Comparison
Identified
Speaker
Agrawal et al.
Proceedings of Meetings on Acoustics, Vol. 19, 060019 (2013) Page 2

---

Preparing HTK Related Text Files:
First a grammar file is made which describes the words to be 
recognized.
The grammar file:
$WORDS = Alpha | Bravo | Charlie | Delta |Echo| Foxtrot | Golf | Hotel | India | Kilo | Lima |Mike |
November | Oscar | Papa | Romeo | Sierra |Tango | Uniform | Victor | Whiskey | Yankee | Zulu;
(sent-start($WORDS)sent-end)
A total of 23 NATO words have used to create a template for each speaker. Then, make the word network for
these words, which take gram.txt as input and generate the word network file “wdnet” that contains each word-
to-work transition.
In order to deal with 23-NATO words only, first we create file called “wlist” that contains only the words in
the prompts file, which can be done by “HTK perl” script.
For building the list of words which exist in the vocabulary, the command HDMan –m –w wlist –n