ALIZE ivector demo GMM UBM

Softwares:

Reference:
http://www1.i2r.a-star.edu.sg/~alarcher/Softwares.html

Tutorial on GMM / UBM for ALIZE 3.0 (download)

In this archive you will find everything required to create your own GMM/UBM system for speaker recognition using ALIZE toolkit. The aim of this demo is to provide an example how to combine ALIZE components in order to learn your Universal Background Model, adapt speaker dependent GMMs, test against speech utterances and normalize the scores.

Config files are just examples and have to be modified in order to produce state-of-the-art-performance. I would like to thank LDC and the NIST for allowing the use of NIST-SRE 2004 data in this tutorial. SRE04 was created by NIST from data and annotations provided by LDC. Sponsors and participants of the NIST Speaker Recognition Evaluations have also shaped the corpus through their requirements, analysis and feedback.

# The following spro feature extraction based GMM UBM framework:
date
bash 01_RUN_feature_extraction.sh > output1.txt
date
bash 02a_RUN_spro_front-end.sh > output2.txt
date
bash 03_RUN_gmm-ubm.sh > output3.txt
date



'via Blog this'