{"url":"/sota/speech-separation-on-whamr","task":{"name":"Speech Separation","url":"/task/speech-separation","note":null},"dataset":{"name":"WHAMR!","url":"/dataset/whamr"},"category":"Speech","categories":["Speech"],"category_note":null,"description":"The task of extracting all overlapping speech sources in a given mixed speech signal refers to the **Speech Separation**. Speech Separation is a special scenario of source separation problem, where the focus is only on the overlapping speech signal sources and other interferences such as music or noise signals are not the main concern of the study. A recent representative Github project can be referred to [ClearerVoice-Studio](https://github.com/modelscope/ClearerVoice-Studio).\r\n\r\n\r\n<span class=\"description-source\">Source: [A Unified Framework for Speech Separation ](https://arxiv.org/abs/1912.07814)</span>\r\n\r\nImage credit: [Speech Separation of A Target Speaker Based on Deep Neural Networks](http://staff.ustc.edu.cn/~jundu/Publications/publications/ICSP2014_Du.pdf)","description_from":"task","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","rank":"the archive's row order at snapshot; not re-ranked","rows_end_at":"2025-07-28","rows_withheld_as_spam":0,"metric_values":"the archive's strings, untouched"},"metrics":["SI-SDRi","MACs (G)","Number of parameters (M)","SDRi"],"metric_direction":{"note":"inferred from the metric name only (the archive records no direction); null = not inferred, chart draws points only","by_metric":{"SI-SDRi":null,"MACs (G)":null,"Number of parameters (M)":null,"SDRi":null}},"counts":{"rows":18,"rows_with_code":15,"rows_with_paper_page":18,"rows_dated":18,"rows_using_additional_data":0},"rows":[{"rank_in_archive_order":1,"model":"TF-Locoformer (M)","metrics":{"Number of parameters (M)":"15","SDRi":"16.9","SI-SDRi":"18.5"},"uses_additional_data":false,"paper_date":"2024-08-06","paper":"/paper/tf-locoformer-transformer-with-local-modeling","paper_url":"https://arxiv.org/abs/2408.03440v1","paper_title":"TF-Locoformer: Transformer with Local Modeling by Convolution for Speech Separation and Enhancement","code":"https://github.com/merlresearch/tf-locoformer","n_code_links":1,"syntology":null},{"rank_in_archive_order":2,"model":"TF-Locoformer (S)","metrics":{"Number of parameters (M)":"5","SDRi":"15.9","SI-SDRi":"17.4"},"uses_additional_data":false,"paper_date":"2024-08-06","paper":"/paper/tf-locoformer-transformer-with-local-modeling","paper_url":"https://arxiv.org/abs/2408.03440v1","paper_title":"TF-Locoformer: Transformer with Local Modeling by Convolution for Speech Separation and Enhancement","code":"https://github.com/merlresearch/tf-locoformer","n_code_links":1,"syntology":null},{"rank_in_archive_order":3,"model":"SepReformer-L + DM","metrics":{"SI-SDRi":"17.1"},"uses_additional_data":false,"paper_date":"2024-06-10","paper":"/paper/separate-and-reconstruct-asymmetric-encoder","paper_url":"https://arxiv.org/abs/2406.05983v4","paper_title":"Separate and Reconstruct: Asymmetric Encoder-Decoder for Speech Separation","code":"https://github.com/dmlguq456/SepReformer","n_code_links":1,"syntology":null},{"rank_in_archive_order":4,"model":"MossFormer2","metrics":{"SI-SDRi":"17.0"},"uses_additional_data":false,"paper_date":"2023-12-19","paper":"/paper/mossformer2-combining-transformer-and-rnn-1","paper_url":"https://arxiv.org/abs/2312.11825v2","paper_title":"MossFormer2: Combining Transformer and RNN-Free Recurrent Network for Enhanced Time-Domain Monaural Speech Separation","code":"https://github.com/modelscope/ClearerVoice-Studio","n_code_links":2,"syntology":{"n_ran":4,"n_unverified":0,"n_samples":4,"n_pointer_only_licence":0}},{"rank_in_archive_order":5,"model":"MossFormer (L) + DM","metrics":{"SI-SDRi":"16.3"},"uses_additional_data":false,"paper_date":"2023-02-23","paper":"/paper/mossformer-pushing-the-performance-limit-of","paper_url":"https://arxiv.org/abs/2302.11824v1","paper_title":"MossFormer: Pushing the Performance Limit of Monaural Speech Separation using Gated Single-Head Transformer with Convolution-Augmented Joint Self-Attentions","code":"https://github.com/modelscope/ClearerVoice-Studio","n_code_links":2,"syntology":{"n_ran":1,"n_unverified":0,"n_samples":1,"n_pointer_only_licence":1}},{"rank_in_archive_order":6,"model":"TD-Conformer (XL) + DM","metrics":{"SI-SDRi":"14.6"},"uses_additional_data":false,"paper_date":"2023-10-09","paper":"/paper/on-time-domain-conformer-models-for-monaural","paper_url":"https://arxiv.org/abs/2310.06125v1","paper_title":"On Time Domain Conformer Models for Monaural Speech Separation in Noisy Reverberant Acoustic Environments","code":"https://github.com/jwr1995/pubsep","n_code_links":1,"syntology":null},{"rank_in_archive_order":7,"model":"Improved Sudo rm -rf (U=36)","metrics":{"SI-SDRi":"13.5"},"uses_additional_data":false,"paper_date":"2021-03-03","paper":"/paper/compute-and-memory-efficient-universal-sound","paper_url":"https://arxiv.org/abs/2103.02644v2","paper_title":"Compute and memory efficient universal sound source separation","code":"https://github.com/etzinis/sudo_rm_rf","n_code_links":3,"syntology":null},{"rank_in_archive_order":8,"model":"TD-Conformer (L) + DM","metrics":{"SI-SDRi":"13.4"},"uses_additional_data":false,"paper_date":"2023-10-09","paper":"/paper/on-time-domain-conformer-models-for-monaural","paper_url":"https://arxiv.org/abs/2310.06125v1","paper_title":"On Time Domain Conformer Models for Monaural Speech Separation in Noisy Reverberant Acoustic Environments","code":"https://github.com/jwr1995/pubsep","n_code_links":1,"syntology":null},{"rank_in_archive_order":9,"model":"Wavesplit","metrics":{"SI-SDRi":"13.2"},"uses_additional_data":false,"paper_date":"2020-02-20","paper":"/paper/wavesplit-end-to-end-speech-separation-by","paper_url":"https://arxiv.org/abs/2002.08933v2","paper_title":"Wavesplit: End-to-End Speech Separation by Speaker Clustering","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":10,"model":"DPTNET - SRSSN","metrics":{"SI-SDRi":"12.3"},"uses_additional_data":false,"paper_date":"2021-10-10","paper":"/paper/stepwise-refining-speech-separation-network","paper_url":"https://arxiv.org/abs/2110.04791v2","paper_title":"Stepwise-Refining Speech Separation Network via Fine-Grained Encoding in High-order Latent Domain","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":11,"model":"DPRNN - SRSSN","metrics":{"SI-SDRi":"12.3"},"uses_additional_data":false,"paper_date":"2021-10-10","paper":"/paper/stepwise-refining-speech-separation-network","paper_url":"https://arxiv.org/abs/2110.04791v2","paper_title":"Stepwise-Refining Speech Separation Network via Fine-Grained Encoding in High-order Latent Domain","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":12,"model":"VSUNOS","metrics":{"SI-SDRi":"12.2"},"uses_additional_data":false,"paper_date":"2020-02-29","paper":"/paper/voice-separation-with-an-unknown-number-of","paper_url":"https://arxiv.org/abs/2003.01531v4","paper_title":"Voice Separation with an Unknown Number of Multiple Speakers","code":"https://github.com/facebookresearch/svoice","n_code_links":4,"syntology":{"n_ran":12,"n_unverified":1,"n_samples":13,"n_pointer_only_licence":9}},{"rank_in_archive_order":13,"model":"Sudo rm -rf (U=16)","metrics":{"SI-SDRi":"12.1"},"uses_additional_data":false,"paper_date":"2020-07-14","paper":"/paper/sudo-rm-rf-efficient-networks-for-universal","paper_url":"https://arxiv.org/abs/2007.06833v1","paper_title":"Sudo rm -rf: Efficient Networks for Universal Audio Source Separation","code":"https://github.com/mpariente/asteroid","n_code_links":4,"syntology":{"n_ran":1,"n_unverified":9,"n_samples":10,"n_pointer_only_licence":0}},{"rank_in_archive_order":14,"model":"TD-Confomer (M) + DM","metrics":{"SI-SDRi":"12"},"uses_additional_data":false,"paper_date":"2023-10-09","paper":"/paper/on-time-domain-conformer-models-for-monaural","paper_url":"https://arxiv.org/abs/2310.06125v1","paper_title":"On Time Domain Conformer Models for Monaural Speech Separation in Noisy Reverberant Acoustic Environments","code":"https://github.com/jwr1995/pubsep","n_code_links":1,"syntology":null},{"rank_in_archive_order":15,"model":"Deformable TCN + Dynamic Mixing","metrics":{"MACs (G)":"3.7","Number of parameters (M)":"3.6","SDRi":"10.3","SI-SDRi":"11.1"},"uses_additional_data":false,"paper_date":"2022-10-27","paper":"/paper/deformable-temporal-convolutional-networks","paper_url":"https://arxiv.org/abs/2210.15305v3","paper_title":"Deformable Temporal Convolutional Networks for Monaural Noisy Reverberant Speech Separation","code":"https://github.com/jwr1995/dtcn","n_code_links":2,"syntology":null},{"rank_in_archive_order":16,"model":"TD-Confomer (S)","metrics":{"SI-SDRi":"10.5"},"uses_additional_data":false,"paper_date":"2023-10-09","paper":"/paper/on-time-domain-conformer-models-for-monaural","paper_url":"https://arxiv.org/abs/2310.06125v1","paper_title":"On Time Domain Conformer Models for Monaural Speech Separation in Noisy Reverberant Acoustic Environments","code":"https://github.com/jwr1995/pubsep","n_code_links":1,"syntology":null},{"rank_in_archive_order":17,"model":"Deformable TCN + Shared Weights + Dynamic Mixing","metrics":{"MACs (G)":"3.7","Number of parameters (M)":"1.3","SDRi":"9.5","SI-SDRi":"10.1"},"uses_additional_data":false,"paper_date":"2022-10-27","paper":"/paper/deformable-temporal-convolutional-networks","paper_url":"https://arxiv.org/abs/2210.15305v3","paper_title":"Deformable Temporal Convolutional Networks for Monaural Noisy Reverberant Speech Separation","code":"https://github.com/jwr1995/dtcn","n_code_links":2,"syntology":null},{"rank_in_archive_order":18,"model":"Bi-LSTM-TASNET","metrics":{"SI-SDRi":"9.2"},"uses_additional_data":false,"paper_date":"2019-07-02","paper":"/paper/wham-extending-speech-separation-to-noisy","paper_url":"https://arxiv.org/abs/1907.01160v1","paper_title":"WHAM!: Extending Speech Separation to Noisy Environments","code":"https://github.com/AkojimaSLP/Neural-mask-estimation","n_code_links":1,"syntology":null}],"since_archive":{"claim":"Results that newer papers report for their own method, placed here by Syntology. A model pointed at the cell in the paper's own table; the number was read from that cell and checked against this leaderboard's metric, dataset, split and scale; an independent check that saw this leaderboard's other rows and every other leaderboard on the same dataset accepted it. Not reviewed by the paper's authors or by the archive's editors, and not ranked against the archive rows.","extraction_file_present":true,"measurement":{"test_papers":883,"papers_with_output":881,"judged_true":108,"judged":110,"wilson95_lower":0.9361,"measured_on":"2026-09-24","frozen_commit":"0e3de0df94"},"measurement_note":"blind adjudication of accepted entries on a held-out split of archive papers, rules frozen before the test","coverage":{"sentence":"Syntology has checked 6,885 of the 9,623 papers on this site that are newer than the archive; results from the others appear after they are checked.","complete":false,"papers_newer_than_archive":9623,"papers_checked":6885,"papers_extracted_not_yet_verified":0,"boards_without_verdict":2,"papers_not_yet_extracted":2737},"order":"newest first by month (arXiv date, else the arXiv-id month), then arXiv id descending","columns":[],"entries":[]},"syntology":{"read_at":"2026-09-25T09:33:49+00:00","claim":"Per row: N of M harvested code samples from that row's paper executed on a synthesized fixture; the other M-N are unverified. Not a reproduction of the row's number; not a correctness claim. n_pointer_only_licence counts samples the site points at rather than redistributes (a licence axis, independent of ran/unverified).","rows_with_graph_line":4,"rows_with_any_sample_ran":4,"distinct_papers_with_graph_line":4,"distinct_papers_with_any_sample_ran":4,"samples_over_distinct_papers":{"n_ran":18,"n_unverified":10,"n_samples":28,"n_pointer_only_licence":10,"note":"each paper (arXiv id) counted once, however many rows it is behind; this is the page-level figure"},"samples_row_weighted":{"n_ran":18,"n_unverified":10,"n_samples":28,"n_pointer_only_licence":10,"note":"row-weighted: a paper behind several rows is counted once per row; inflated relative to samples_over_distinct_papers by design, kept for readers summing the per-row syntology blocks"}}}