FUNDAMENTALS OF SPEAKER RECOGNITION :- UMANG PATEL (ujp2001)

PROJECT NAME :- SPEKAER IDENTIFICATION USING I-VECTORS Contact ujp2001@columbia.edu

Command to run the DEMO: features.m final_run.m

Dependencies required for this project to RUN:

VOICEBOX : Speech Processing Toolbox for MATLAB
http://www.ee.ic.ac.uk/hp/staff/dmb/voicebox/voicebox.html

The files included are :

features.m

• To extract features from audio files.

ubm_train.m

• To train the UBM model on training and enrolment data.

ubm.m

• To train the total Variability matrix V in "s=m +Vy"

LDA.m

• To run LDA

HLDA.m

• To run HLDA

GMM_baseline.m

• To run GMM baseline

final_run.m

• To run the entire program (intenrally calls all the above mentioned file)

figure_gen.m

• To generate all the figures

cofactor.m

• To find cofactor of matrix (internally called by HLDA.m)

CDS.m

• To compute the CDS

Main instructions:

features.m
1)Before caling final_run.m call the features.m so that it will create appropriate feature matrix

final_run.m
1) It calls the above mentioned files internally.
2) Change the num_gauss(line 11 to change number of GMM models) ,d(line 13 to change the dimension of i-vectors),trainingindex(line 16 to specify the proportion of training features from each folder)
3) Change the above mentioned parameers to get different results.

GMM_baseline.m
1)After running final_run.m run GMM_baseline.m because it used 'g_ubm'(UBM Model Parameters).