eprintid: 10154106 rev_number: 11 eprint_status: archive userid: 699 dir: disk0/10/15/41/06 datestamp: 2022-08-23 10:56:53 lastmod: 2022-08-23 10:56:53 status_changed: 2022-08-23 10:56:53 type: proceedings_section metadata_visibility: show sword_depositor: 699 creators_name: Wang, J creators_name: Lam, MWY creators_name: Su, D creators_name: Yu, D title: Contrastive separative coding for self-supervised representation learning ispublished: pub divisions: C05 divisions: F48 divisions: B04 divisions: UCL keywords: Speaker verification, speech separation, self attention, contrastive loss note: This version is the author accepted manuscript. For information on re-use, please refer to the publisher's terms and conditions. abstract: To extract robust deep representations from long sequential modeling of speech data, we propose a self-supervised learning approach, namely Contrastive Separative Coding (CSC). Our key finding is to learn such representations by separating the target signal from contrastive interfering signals. First, a multi-task separative encoder is built to extract shared separable and discriminative embedding; secondly, we propose a powerful cross-attention mechanism performed over speaker representations across various interfering conditions, allowing the model to focus on and globally aggregate the most critical information to answer the "query"(current bottom-up embedding) while paying less attention to interfering, noisy, or irrelevant parts; lastly, we form a new probabilistic contrastive loss which estimates and maximizes the mutual information between the representations and the global speaker vector. While most prior unsupervised methods have focused on predicting the future, neighboring, or missing samples, we take a different perspective of predicting the interfered samples. Moreover, our contrastive separative loss is free from negative sampling. The experiment demonstrates that our approach can learn useful representations achieving a strong speaker verification performance in adverse conditions. date: 2021 date_type: published publisher: IEEE official_url: https://doi.org/10.1109/ICASSP39728.2021.9414352 oa_status: green full_text_type: other language: eng primo: open primo_central: open_green verified: verified_manual elements_id: 1971130 doi: 10.1109/ICASSP39728.2021.9414352 isbn_13: 978-1-7281-7605-5 lyricists_name: Wang, Jun lyricists_id: JWANG00 actors_name: Flynn, Bernadette actors_id: BFFLY94 actors_role: owner full_text_status: public pres_type: paper publication: ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings pagerange: 3865-3869 event_title: ICASSP - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) event_location: Toronto, ON, Canada event_dates: 6th-11th June 2021 book_title: Proceedings of the ICASSP - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) citation: Wang, J; Lam, MWY; Su, D; Yu, D; (2021) Contrastive separative coding for self-supervised representation learning. In: Proceedings of the ICASSP - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). (pp. pp. 3865-3869). IEEE Green open access document_url: https://discovery.ucl.ac.uk/id/eprint/10154106/1/2103.00816.pdf