Přístupnostní navigace
E-application
Search Search Close
Publication detail
WANG, S. ROHDIN, J. PLCHOT, O. BURGET, L. YU, K. ČERNOCKÝ, J.
Original Title
Investigation of Specaugment for Deep Speaker Embedding Learning
Type
conference paper
Language
English
Original Abstract
SpecAugment is a newly proposed data augmentation method for speech recognition. By randomly masking bands in the log Mel spectogram this method leads to impressive performance improvements. In this paper, we investigate the usage of SpecAugment for speaker verification tasks. Two different models, namely 1-D convolutional TDNN and 2-D convolutional ResNet34, trained with either Softmax or AAM-Softmax loss, are used to analyze SpecAugments effectiveness. Experiments are carried out on the Voxceleb and NIST SRE 2016 dataset. By applying SpecAugment to the original clean data in an on-the-fly manner without complex off-line data augmentation methods, we obtained 3.72% and 11.49% EER for NIST SRE 2016 Cantonese and Tagalog, respectively. For Voxceleb1 evaluation set, we obtained 1.47% EER.
Keywords
speaker embedding, on-the-fly data augmentation, speaker verification, specaugment
Authors
WANG, S.; ROHDIN, J.; PLCHOT, O.; BURGET, L.; YU, K.; ČERNOCKÝ, J.
Released
4. 5. 2020
Publisher
IEEE Signal Processing Society
Location
Barcelona
ISBN
978-1-5090-6631-5
Book
ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings
Pages from
7139
Pages to
7143
Pages count
5
URL
https://ieeexplore.ieee.org/document/9053481/authors#authors
BibTex
@inproceedings{BUT163947, author="WANG, S. and ROHDIN, J. and PLCHOT, O. and BURGET, L. and YU, K. and ČERNOCKÝ, J.", title="Investigation of Specaugment for Deep Speaker Embedding Learning", booktitle="ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings", year="2020", pages="7139--7143", publisher="IEEE Signal Processing Society", address="Barcelona", doi="10.1109/ICASSP40776.2020.9053481", isbn="978-1-5090-6631-5", url="https://ieeexplore.ieee.org/document/9053481/authors#authors" }