{"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/single-channel-multi-speaker-separation-using","title":"Single-Channel Multi-Speaker Separation using Deep Clustering","arxiv_id":"1607.02173","date":"2016-07-07","proceeding":null,"authors":["Yusuf Isik","Jonathan Le Roux","Zhuo Chen","Shinji Watanabe","John R. Hershey"],"abstract":"Deep clustering is a recently introduced deep learning architecture that uses\ndiscriminatively trained embeddings as the basis for clustering. It was\nrecently applied to spectrogram segmentation, resulting in impressive results\non speaker-independent multi-speaker separation. In this paper we extend the\nbaseline system with an end-to-end signal approximation objective that greatly\nimproves performance on a challenging speech separation. We first significantly\nimprove upon the baseline system performance by incorporating better\nregularization, larger temporal context, and a deeper architecture, culminating\nin an overall improvement in signal to distortion ratio (SDR) of 10.3 dB\ncompared to the baseline of 6.0 dB for two-speaker separation, as well as a 7.1\ndB SDR improvement for three-speaker separation. We then extend the model to\nincorporate an enhancement layer to refine the signal estimates, and perform\nend-to-end training through both the clustering and enhancement stages to\nmaximize signal fidelity. We evaluate the results using automatic speech\nrecognition. The new signal approximation objective, combined with end-to-end\ntraining, produces unprecedented performance, reducing the word error rate\n(WER) from 89.1% down to 30.8%. This represents a major advancement towards\nsolving the cocktail party problem.","url_abs":"http://arxiv.org/abs/1607.02173v1","url_pdf":"http://arxiv.org/pdf/1607.02173v1.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":"single-channel-multi-speaker-separation-using","repo_url":"https://github.com/ishandutta2007/Speech-Denoising-Landscape","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"unanswered"}},{"paper_slug":"single-channel-multi-speaker-separation-using","repo_url":"https://github.com/JusperLee/Deep-Clustering-for-Speech-Separation","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":{"status":"ok"}}],"tasks":[{"task_slug":"automatic-speech-recognition-2","task_name":"Automatic Speech Recognition"},{"task_slug":"automatic-speech-recognition","task_name":"Automatic Speech Recognition (ASR)"},{"task_slug":"clustering","task_name":"Clustering"},{"task_slug":"deep-clustering","task_name":"Deep Clustering"},{"task_slug":"speaker-separation","task_name":"Speaker Separation"},{"task_slug":"speech-recognition","task_name":"Speech Recognition"},{"task_slug":"speech-separation","task_name":"Speech Separation"},{"task_slug":"speech-recognition-1","task_name":"speech-recognition"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/1607.02173","atlas_url":"https://app.syntology.ai/?focus=1607.02173","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}