Summary
The 2018 International Symposium on Information Theory and Its Applications　(ISITA2018)
2018
Session Number:We-AM-Poster
Session:
Number:We-AM-Poster.8
A Thresholded Discriminative Metric Learning Approach for Deep Speaker Recognition
Yin-Cheng Yeh, Yan-Chin Liao, Hsie-Chia Chang,
pp.498-498
Publication Date:2018/10/18
Online ISSN:2188-5079
DOI:10.34385/proc.55.We-AM-Poster.8
PDF download
Summary:
In this paper, we present a deep-learning model for speaker recognition. To enhance text-independent open-set verification and identification, a discriminative metric learning method with thresholded center loss function is proposed. Our method mitigates the performance gap between the models trained from a studio-quality dataset and that from a noisy dataset.