{"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/exploring-the-encoding-layer-and-loss","title":"Exploring the Encoding Layer and Loss Function in End-to-End Speaker and Language Recognition System","arxiv_id":"1804.05160","date":"2018-04-14","proceeding":null,"authors":["Weicheng Cai","Jinkun Chen","Ming Li"],"abstract":"In this paper, we explore the encoding/pooling layer and loss function in the\nend-to-end speaker and language recognition system. First, a unified and\ninterpretable end-to-end system for both speaker and language recognition is\ndeveloped. It accepts variable-length input and produces an utterance level\nresult. In the end-to-end system, the encoding layer plays a role in\naggregating the variable-length input sequence into an utterance level\nrepresentation. Besides the basic temporal average pooling, we introduce a\nself-attentive pooling layer and a learnable dictionary encoding layer to get\nthe utterance level representation. In terms of loss function for open-set\nspeaker verification, to get more discriminative speaker embedding, center loss\nand angular softmax loss is introduced in the end-to-end system. Experimental\nresults on Voxceleb and NIST LRE 07 datasets show that the performance of\nend-to-end learning system could be significantly improved by the proposed\nencoding layer and loss function.","url_abs":"http://arxiv.org/abs/1804.05160v1","url_pdf":"http://arxiv.org/pdf/1804.05160v1.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":"exploring-the-encoding-layer-and-loss","repo_url":"https://github.com/jkchen79/netvlad-in-speech","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}}],"tasks":[{"task_slug":"speaker-verification","task_name":"Speaker Verification"}],"methods":[{"method_slug":"softmax","method_name":"Softmax"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/1804.05160","atlas_url":"https://app.syntology.ai/?focus=1804.05160","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}