eprintid: 10154105 rev_number: 13 eprint_status: archive userid: 699 dir: disk0/10/15/41/05 datestamp: 2022-08-23 11:04:17 lastmod: 2022-08-23 11:04:17 status_changed: 2022-08-23 11:04:17 type: proceedings_section metadata_visibility: show sword_depositor: 699 creators_name: Lam, MWY creators_name: Wang, J creators_name: Su, D creators_name: Yu, D title: Effective Low-Cost Time-Domain Audio Separation Using Globally Attentive Locally Recurrent Networks ispublished: pub divisions: C05 divisions: F48 divisions: B04 divisions: UCL keywords: speech separation, TasNet, low-cost, multi-head attention note: This version is the author accepted manuscript. For information on re-use, please refer to the publisher's terms and conditions. abstract: Recent research on the time-domain audio separation networks (TasNets) has brought great success to speech separation. Nevertheless, conventional TasNets struggle to satisfy the memory and latency constraints in industrial applications. In this regard, we design a low-cost high-performance architecture, namely, globally attentive locally recurrent (GALR) network. Alike the dual-path RNN (DPRNN), we first split a feature sequence into 2D segments and then process the sequence along both the intra- and inter-segment dimensions. Our main innovation lies in that, on top of features recurrently processed along the inter-segment dimensions, GALR applies a self-attention mechanism to the sequence along the inter-segment dimension, which aggregates context-aware information and also enables parallelization. Our experiments suggest that GALR is a notably more effective network than the prior work. On one hand, with only 1.5M parameters, it has achieved comparable separation performance at a much lower cost with 36.1% less runtime memory and 49.4% fewer computational operations, relative to the DPRNN. On the other hand, in a comparable model size with DPRNN, GALR has consistently outperformed DPRNN in three datasets, in particular, with a substantial margin of 2.4dB absolute improvement of SI-SNRi in the benchmark WSJ0-2mix task. date: 2021 date_type: published publisher: IEEE official_url: https://doi.org/10.1109/SLT48900.2021.9383464 oa_status: green full_text_type: other language: eng primo: open primo_central: open_green verified: verified_manual elements_id: 1971132 doi: 10.1109/SLT48900.2021.9383464 isbn_13: 9781728170664 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: 2021 IEEE Spoken Language Technology Workshop, SLT 2021 - Proceedings pagerange: 801-808 event_title: 2021 IEEE Spoken Language Technology Workshop (SLT) event_location: Shenzhen, China event_dates: 19th-22nd Jan 2021 book_title: Proceedings of the 2021 IEEE Spoken Language Technology Workshop (SLT) citation: Lam, MWY; Wang, J; Su, D; Yu, D; (2021) Effective Low-Cost Time-Domain Audio Separation Using Globally Attentive Locally Recurrent Networks. In: Proceedings of the 2021 IEEE Spoken Language Technology Workshop (SLT). (pp. pp. 801-808). IEEE Green open access document_url: https://discovery.ucl.ac.uk/id/eprint/10154105/1/2101.05014.pdf