Dependent on
============
1. HTK (please install and make sure it is working)
2. netlab (provided with the SII software)

USER input:
===========
0. Use HTK and create a recognizer using single-state HMM (i.e., transition matrix is 3x3) from a suitable corpus (e.g., TIMIT) and prepare the followings: 
	- provide models/ folder containing hmmdefs
    	- HTK configuration file as htk_config
	- phone list (phones) and dictionary (dico) [dico is nothing but same phoneme list in two columns]
	- word net (wdnet)
1. Files from the training subject (at least need to have 400-500 sentences for good acoustic-to-articulatory inversion)
	- an scp file containing the list of mfc files of the training subject - [arti_train.scp] (at a rate of 100Hz)
	- a control file containing the list of the .mat filenames (with full path) of the target probability feature files (to be generated by SII code) corresponding to each file in arti_train.scp - [arti_train.pftr] (e.g., if i-th line in arti_train.scp has abc/xyz/edf.mfc then the i-th line in arti_train.pftr should be def/yhk/edf.mat) 
	- a control file containing the list of articulatory (.mat) files - [arti_train.arti] (each articulatory mat file should contain Nxd matrix representing d-dimensional articulatory features in N frames] The sampling rate of articulatory data should be 100Hz.
 	- provide classes and xforms_train directory from adaptation of hmmdefs(HTK HMM model) using the training subject data(for details on xforms please read HTK manual) **
2. Files from the test subject (at least need to have 20 sentences for good adaptation)
	- an scp file containing the list of mfc files of the training subject - [arti_test.scp] (at a rate of 100Hz)
	- a control file containing the list of the .mat filenames (with full path) of the target probability feature files (to be generated by SII code) corresponding to each file in arti_test.scp - [arti_test.pftr] (e.g., if i-th line in arti_test.scp has abc/xyz/edf1.mfc then the i-th line in arti_test.pftr should be def/yhk/edf1.mat)
	- a control file containing the list of target estimated articulatory (.mat) files - [arti_test.arti]
 	- provide xforms_test directory from adaptation of hmmdefs (HTK HMM model) using the test subject data(for details on xforms please read HTK manual) **

STEPS:
======
-. go in SII_wa_v0
-. copy the models/ folder in current location.
-. mkdir -p GAS_scheme2;rm -rf GAS_scheme2/* (change according to your OS)
-. open MATLAB and run 
	>> create_phn_gmms('models/hmmdefs','GAS_scheme2/');
-. mkdir control_files/
-. in MATLAB run (change according to your OS)
	>> create_mfc_train_dev('arti_train.scp'); %% give the full path of arti_train.scp if it is not in the current directory
	>> create_pftr_train_dev('arti_train.pftr'); %% give the full path of arti_train.pftr if it is not in the current directory
	>> create_arti_train_dev('arti_train.arti'); %% give the full path of arti_train.arti if it is not in the current directory
-. copy arti_test.scp into control_files/test_mfc.txt
-. copy arti_test.pftr into control_files/test_pftr.txt
-. copy arti_test.arti into control_files/test_arti.txt
-. mkdir -p label_files;rm -rf label_files/* ; mkdir label_files/train ; mkdir label_files/dev ; mkdir label_files/test ;(change according to your OS)
-. in MATLAB run
    >> create_LUT('phones')
-. run scheme2_sii_wa



**For adaptation (optional)
=========================
Other files required

-.regtree.hed
-.SPKRlabel - speaker files MLF
-.SPKRfiles - speaker files list
-.config.global
-.config.rc

run doadaptation


