Loading...
Loading...

Go to the content (press return)

From features to speaker vectors by means of restricted Boltzmann machine adaptation

Author
Safari, P.; Ghahabi, O.; Hernando, J.
Type of activity
Presentation of work at congresses
Name of edition
ODYSSEY 2016 - The Speaker and Language Recognition Workshop
Date of publication
2016
Presentation's date
2016-06-24
Book of congress proceedings
ODYSSEY 2016 - The Speaker and Language Recognition Workshop
First page
366
Last page
371
DOI
https://doi.org/10.21437/Odyssey.2016-53 Open in new window
Repository
http://hdl.handle.net/2117/101682 Open in new window
URL
http://www.isca-speech.org/archive/Odyssey_2016/pdfs/15.pdf Open in new window
Abstract
Restricted Boltzmann Machines (RBMs) have shown success in different stages of speaker recognition systems. In this paper, we propose a novel framework to produce a vector-based representation for each speaker, which will be referred to as RBM-vector. This new approach maps the speaker spectral features to a single fixed-dimensional vector carrying speaker-specific information. In this work, a global model, referred to as Universal RBM (URBM), is trained taking advantage of RBM unsupervised lear...
Citation
Safari, P., Ghahabi, O., Hernando, J. From features to speaker vectors by means of restricted Boltzmann machine adaptation. A: The Speaker and Language Recognition Workshop. "ODYSSEY 2016 - The Speaker and Language Recognition Workshop". Bilbao: 2016, p. 366-371.
Keywords
RBM-vector, RBMs, Restricted Boltzmann Machines, Speaker recognition
Group of research
TALP - Centre for Language and Speech Technologies and Applications
Universitat Politècnica de Catalunya
VEU - Speech Processing Group

Participants

Attachments