Sorry, you need to enable JavaScript to visit this website.

facebooktwittermailshare

GENERALISED DISCRIMINATIVE TRANSFORM VIA CURRICULUM LEARNING FOR SPEAKER RECOGNITION

Abstract: 

In this paper we introduce a speaker verification system deployed on mobile devices that can be used to personalise a keyword spotter. We describe a baseline DNN system that maps an utterance to a speaker embedding, which is used to measure speaker differences via cosine similarity. We then introduce an architectural modification which uses an LSTM system where the parameters are optimised via a curriculum learning procedure to reduce the detection error and improve its generalisability across various conditions. Experiments on our internal datasets show that the proposed approach outperforms the DNN baseline system and yields a relative EER reduction of 30–70% on both text-dependent and text-independent tasks under a variety of acoustic conditions.

up
0 users have voted:

Paper Details

Authors:
Erik Marchi, Stephen Shum, Kyuyeon Hwang, Sachin Kajarekar, Siddharth Sigtia, Hywel Richards, Rob Haynes, Yoon Kim, John Bridle
Submitted On:
23 April 2018 - 1:24am
Short Link:
Type:
Poster
Event:
Presenter's Name:
Erik Marchi
Paper Code:
SP-P7.1
Document Year:
2018
Cite

Document Files

Siri_PHS_CurriculumLearning_ICASSP18v3.pdf

(161 downloads)

Subscribe

[1] Erik Marchi, Stephen Shum, Kyuyeon Hwang, Sachin Kajarekar, Siddharth Sigtia, Hywel Richards, Rob Haynes, Yoon Kim, John Bridle, "GENERALISED DISCRIMINATIVE TRANSFORM VIA CURRICULUM LEARNING FOR SPEAKER RECOGNITION", IEEE SigPort, 2018. [Online]. Available: http://sigport.org/3144. Accessed: Aug. 18, 2018.
@article{3144-18,
url = {http://sigport.org/3144},
author = {Erik Marchi; Stephen Shum; Kyuyeon Hwang; Sachin Kajarekar; Siddharth Sigtia; Hywel Richards; Rob Haynes; Yoon Kim; John Bridle },
publisher = {IEEE SigPort},
title = {GENERALISED DISCRIMINATIVE TRANSFORM VIA CURRICULUM LEARNING FOR SPEAKER RECOGNITION},
year = {2018} }
TY - EJOUR
T1 - GENERALISED DISCRIMINATIVE TRANSFORM VIA CURRICULUM LEARNING FOR SPEAKER RECOGNITION
AU - Erik Marchi; Stephen Shum; Kyuyeon Hwang; Sachin Kajarekar; Siddharth Sigtia; Hywel Richards; Rob Haynes; Yoon Kim; John Bridle
PY - 2018
PB - IEEE SigPort
UR - http://sigport.org/3144
ER -
Erik Marchi, Stephen Shum, Kyuyeon Hwang, Sachin Kajarekar, Siddharth Sigtia, Hywel Richards, Rob Haynes, Yoon Kim, John Bridle. (2018). GENERALISED DISCRIMINATIVE TRANSFORM VIA CURRICULUM LEARNING FOR SPEAKER RECOGNITION. IEEE SigPort. http://sigport.org/3144
Erik Marchi, Stephen Shum, Kyuyeon Hwang, Sachin Kajarekar, Siddharth Sigtia, Hywel Richards, Rob Haynes, Yoon Kim, John Bridle, 2018. GENERALISED DISCRIMINATIVE TRANSFORM VIA CURRICULUM LEARNING FOR SPEAKER RECOGNITION. Available at: http://sigport.org/3144.
Erik Marchi, Stephen Shum, Kyuyeon Hwang, Sachin Kajarekar, Siddharth Sigtia, Hywel Richards, Rob Haynes, Yoon Kim, John Bridle. (2018). "GENERALISED DISCRIMINATIVE TRANSFORM VIA CURRICULUM LEARNING FOR SPEAKER RECOGNITION." Web.
1. Erik Marchi, Stephen Shum, Kyuyeon Hwang, Sachin Kajarekar, Siddharth Sigtia, Hywel Richards, Rob Haynes, Yoon Kim, John Bridle. GENERALISED DISCRIMINATIVE TRANSFORM VIA CURRICULUM LEARNING FOR SPEAKER RECOGNITION [Internet]. IEEE SigPort; 2018. Available from : http://sigport.org/3144