Automatic speaker recognition experiments with deep neural nets This repository contains code to replicate speaker recognition experiments on TIMIT database using DNN
(1) A baseline GMM-UBM implemtation of microsoft toolbox (MST Identity) is also added.
(2) For DNNs, several initialization schemes like (normalized initialization, random walk init.. etc ,) and learning rate methods like (sgd-cm,ada-delta and adam) are implemented in matlab.
It is a GPU based code. :)