https://scholars.lib.ntu.edu.tw/handle/123456789/607155
標題: | One shot learning for speech separation | 作者: | Wu Y.-K Huang K.-P Tsao Y HUNG-YI LEE |
關鍵字: | ANIL;MAML;Meta-learning;Speech separation;Separation;Speech analysis;Feature reuse;Inner loops;Metalearning;Noisy environment;One-shot learning;Task-specific modules;Use-model;Source separation | 公開日期: | 2021 | 卷: | 2021-June | 起(迄)頁: | 5769-5773 | 來源出版物: | ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings | 摘要: | Despite the recent success of speech separation models, they fail to separate sources properly while facing different sets of people or noisy environments. To tackle this problem, we proposed to apply meta-learning to the speech separation task. We aimed to find a meta-initialization model, which can quickly adapt to new speakers by seeing only one mixture generated by those people. In this paper, we use model-agnostic meta-learning(MAML) algorithm and almost no inner loop(ANIL) algorithm in Conv-TasNet to achieve this goal. The experiment results show that our model can adapt not only to a new set of speakers but also noisy environments. Furthermore, we found out that the encoder and decoder serve as the feature-reuse layers, while the separator is the task-specific module. ? 2021 IEEE |
URI: | https://www.scopus.com/inward/record.uri?eid=2-s2.0-85115144481&doi=10.1109%2fICASSP39728.2021.9413956&partnerID=40&md5=3e90b74bb231d719552774e444544ec9 https://scholars.lib.ntu.edu.tw/handle/123456789/607155 |
ISSN: | 15206149 | DOI: | 10.1109/ICASSP39728.2021.9413956 |
顯示於: | 電機工程學系 |
在 IR 系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。