I am working for voice authentication. First, I extracted voice features from user as MFCC. After that, I use the "gmm_estimate" function to get the mean, variance and wights to form the GMM.
For now, let say I have multiple voice samples from same user and I formed GM Model for each of them. After that I am going to train/fit all of them into 1 GM Model and compared with testing voice from user input. What I did so far is use the "fitgmdist" function but failed with this error :"Error using gmcluster (line 197) Ill-conditioned covariance created at iteration 4."
function MFCC_feat_inject(sig,features_mfcc_file,na)
No_of_Gaussians=12;
load(features_mfcc_file);
% no_of_fe will have the no of saved feature
% fe matrix will have the feature
no_of_fe=no_of_fe+1;
LEN=length(na);
name(no_of_fe,1:LEN)=char(na);
fe=melcepst(sig,8000);
[mu_train,sigma_train,c_train]=gmm_estimate(fe(:,5:12)',No_of_Gaussians,20);
%mean
fea{no_of_fe,1}=mu_train;
%variance
fea{no_of_fe,2}=sigma_train;
%weights
fea{no_of_fe,3}=c_train;
save(features_mfcc_file,'no_of_fe','name','fea');
