{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/tasnet-time-domain-audio-separation-network","title":"TasNet: time-domain audio separation network for real-time, single-channel speech separation","arxiv_id":"1711.00541","date":"2017-11-01","proceeding":null,"authors":["Yi Luo","Nima Mesgarani"],"abstract":"Robust speech processing in multi-talker environments requires effective\nspeech separation. Recent deep learning systems have made significant progress\ntoward solving this problem, yet it remains challenging particularly in\nreal-time, short latency applications. Most methods attempt to construct a mask\nfor each source in time-frequency representation of the mixture signal which is\nnot necessarily an optimal representation for speech separation. In addition,\ntime-frequency decomposition results in inherent problems such as\nphase/magnitude decoupling and long time window which is required to achieve\nsufficient frequency resolution. We propose Time-domain Audio Separation\nNetwork (TasNet) to overcome these limitations. We directly model the signal in\nthe time-domain using an encoder-decoder framework and perform the source\nseparation on nonnegative encoder outputs. This method removes the frequency\ndecomposition step and reduces the separation problem to estimation of source\nmasks on encoder outputs which is then synthesized by the decoder. Our system\noutperforms the current state-of-the-art causal and noncausal speech separation\nalgorithms, reduces the computational cost of speech separation, and\nsignificantly reduces the minimum required latency of the output. This makes\nTasNet suitable for applications where low-power, real-time implementation is\ndesirable such as in hearable and telecommunication devices.","url_abs":"http://arxiv.org/abs/1711.00541v2","url_pdf":"http://arxiv.org/pdf/1711.00541v2.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"tasnet-time-domain-audio-separation-network","repo_url":"https://github.com/SsssSunhy/Tasnet-Mindspore-on-Librimix","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"mindspore","reach":{"status":"ok"}},{"paper_slug":"tasnet-time-domain-audio-separation-network","repo_url":"https://github.com/mpariente/asteroid","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}},{"paper_slug":"tasnet-time-domain-audio-separation-network","repo_url":"https://github.com/kaituoxu/TasNet","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":{"status":"ok"}}],"tasks":[{"task_slug":"decoder","task_name":"Decoder"},{"task_slug":"speech-separation","task_name":"Speech Separation"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/speech-separation-on-wsj0-2mix","task":"Speech Separation","dataset":"WSJ0-2mix","model":"TasNet","rank_in_archive_order":39,"of":40,"metrics":{"SI-SDRi":"10.8"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1711.00541","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}