Investigation of DNN based feature enhancement jointly trained with x-vectors for noise-robust speaker verification
- Authors
- Yang, Joon-Young; Park, Kwan-Ho; Chang, Joon-Hyuk; Kim, Youngsam; Cho, Sangrae
- Issue Date
- Jan-2020
- Publisher
- Institute of Electrical and Electronics Engineers Inc.
- Keywords
- Deep speaker embedding; Feature enhancement; Joint training; Speaker verification
- Citation
- 2020 International Conference on Electronics, Information, and Communication, ICEIC 2020, pp.1 - 5
- Indexed
- SCOPUS
- Journal Title
- 2020 International Conference on Electronics, Information, and Communication, ICEIC 2020
- Start Page
- 1
- End Page
- 5
- URI
- https://scholarworks.bwise.kr/hanyang/handle/2021.sw.hanyang/4465
- DOI
- 10.1109/ICEIC49074.2020.9051093
- Abstract
- In this paper, we investigate the deep neural network (DNN) based feature enhancement as the denoising frontend of the x-vector speaker verification framework in noisy environments. Firstly, the feature enhancement DNN (FE-DNN) learns the mapping function from the noisy to the clean corpora on the frame-level acoustic feature domain, and then the x-vector network (XvectorNet) is trained on top of the enhanced features. Finally, the separately trained FE-DNN and the XvectorNet are serially concatenated and jointly trained under the supervision of cross-entropy loss. In addition., we adopt the logistic margin softmax layer for training the XvectorNet in order to obtain more discriminative speaker embeddings.
- Files in This Item
-
Go to Link
- Appears in
Collections - 서울 공과대학 > 서울 융합전자공학부 > 1. Journal Articles
![qrcode](https://api.qrserver.com/v1/create-qr-code/?size=55x55&data=https://scholarworks.bwise.kr/hanyang/handle/2021.sw.hanyang/4465)
Items in ScholarWorks are protected by copyright, with all rights reserved, unless otherwise indicated.