{"links":{},"metadata":{"_oai":{"id":"oai:ipsj.ixsq.nii.ac.jp:00062667","sets":["1164:5159:5160:5697"]},"path":["5697"],"owner":"10","recid":"62667","title":["最小相対エントロピー識別学習に基づくカーネルマシンを利用した音声認識"],"pubdate":{"attribute_name":"公開日","attribute_value":"2009-07-10"},"_buckets":{"deposit":"472e9ad1-78e9-4ffd-a631-7efed3d5567f"},"_deposit":{"id":"62667","pid":{"type":"depid","value":"62667","revision_id":0},"owners":[10],"status":"published","created_by":10},"item_title":"最小相対エントロピー識別学習に基づくカーネルマシンを利用した音声認識","author_link":["0","0"],"item_titles":{"attribute_name":"タイトル","attribute_value_mlt":[{"subitem_title":"最小相対エントロピー識別学習に基づくカーネルマシンを利用した音声認識"},{"subitem_title":"A Kernel Machine Derived by Minimum Relative Entropy Discrimination For Automatic Speech Recognition","subitem_title_language":"en"}]},"item_keyword":{"attribute_name":"キーワード","attribute_value_mlt":[{"subitem_subject":"音響モデル","subitem_subject_scheme":"Other"}]},"item_type_id":"4","publish_date":"2009-07-10","item_4_text_3":{"attribute_name":"著者所属","attribute_value_mlt":[{"subitem_text_value":"早稲田大学"},{"subitem_text_value":"NTTコミュニケーション科学基礎研究所"},{"subitem_text_value":"NTTコミュニケーション科学基礎研究所"},{"subitem_text_value":"NTTコミュニケーション科学基礎研究所"},{"subitem_text_value":"早稲田大学"}]},"item_4_text_4":{"attribute_name":"著者所属(英)","attribute_value_mlt":[{"subitem_text_value":"Waseda University","subitem_text_language":"en"},{"subitem_text_value":"NTT Communication Science Laboratories","subitem_text_language":"en"},{"subitem_text_value":"NTT Communication Science Laboratories","subitem_text_language":"en"},{"subitem_text_value":"NTT Communication Science Laboratories","subitem_text_language":"en"},{"subitem_text_value":"Waseda University","subitem_text_language":"en"}]},"item_language":{"attribute_name":"言語","attribute_value_mlt":[{"subitem_language":"jpn"}]},"item_publisher":{"attribute_name":"出版者","attribute_value_mlt":[{"subitem_publisher":"情報処理学会","subitem_publisher_language":"ja"}]},"publish_status":"0","weko_shared_id":-1,"item_file_price":{"attribute_name":"Billing file","attribute_type":"file","attribute_value_mlt":[{"url":{"url":"https://ipsj.ixsq.nii.ac.jp/record/62667/files/IPSJ-SLP09077006.pdf"},"date":[{"dateType":"Available","dateValue":"2011-07-10"}],"format":"application/pdf","billing":["billing_file"],"filename":"IPSJ-SLP09077006.pdf","filesize":[{"value":"319.0 kB"}],"mimetype":"application/pdf","priceinfo":[{"tax":["include_tax"],"price":"660","billingrole":"5"},{"tax":["include_tax"],"price":"330","billingrole":"6"},{"tax":["include_tax"],"price":"0","billingrole":"22"},{"tax":["include_tax"],"price":"0","billingrole":"44"}],"accessrole":"open_date","version_id":"751b0955-0022-4bca-939d-1f92417607d3","displaytype":"detail","licensetype":"license_note","license_note":"Copyright (c) 2009 by the Information Processing Society of Japan"}]},"item_4_creator_5":{"attribute_name":"著者名","attribute_type":"creator","attribute_value_mlt":[{"creatorNames":[{"creatorName":"久保, 陽太郎"},{"creatorName":"渡部, 晋治"},{"creatorName":"中村, 篤"},{"creatorName":"エリックマクダーモット"},{"creatorName":"小林, 哲則"}],"nameIdentifiers":[{}]}]},"item_4_creator_6":{"attribute_name":"著者名(英)","attribute_type":"creator","attribute_value_mlt":[{"creatorNames":[{"creatorName":"Yotaro, Kubo","creatorNameLang":"en"},{"creatorName":"Shinji, Watanabe","creatorNameLang":"en"},{"creatorName":"Atsushi, Nakamura","creatorNameLang":"en"},{"creatorName":"Erik, McDermott","creatorNameLang":"en"},{"creatorName":"Tetsunori, Kobayashi","creatorNameLang":"en"}],"nameIdentifiers":[{}]}]},"item_4_source_id_9":{"attribute_name":"書誌レコードID","attribute_value_mlt":[{"subitem_source_identifier":"AN10442647","subitem_source_identifier_type":"NCID"}]},"item_4_textarea_12":{"attribute_name":"Notice","attribute_value_mlt":[{"subitem_textarea_value":"SIG Technical Reports are nonrefereed and hence may later appear in any journals, conferences, symposia, etc."}]},"item_resource_type":{"attribute_name":"資源タイプ","attribute_value_mlt":[{"resourceuri":"http://purl.org/coar/resource_type/c_18gh","resourcetype":"technical report"}]},"item_4_description_7":{"attribute_name":"論文抄録","attribute_value_mlt":[{"subitem_description":"本稿ではカーネルマシンに基づく音声認識法を提案する．提案手法では，Log-Linear タイプの出力分布を持つ連続分布型隠れマルコフモデル (CD-HMM) を最小相対エントロピー識別によって学習させる時の目的関数，および学習されたモデルのパラメタにカーネル法に基づく表現を導入する．従来のカーネル法ベースの音声認識と異なり，提案法は隠れマルコフモデルの構造をそのまま利用しているため，音声認識問題をフレーム列，隠れ状態列，ラベル列の 3 つの系列の変換問題として取り扱うことができる．そのため，提案法には動的計画法を用いた各種探索技法を容易に組み込むことができる．これは現代的な音声認識システムの実装には必要不可欠な要素である．加えて，本稿では提案法を効率的に実現するために，Cutting Plane 法を拡張した最適化アルゴリズムについても提案する．提案法の評価は孤立音素認識タスクにて行なった．評価の結果として，提案モデルがテストデータを用いて十分に調整した CD-HMM と同等の性能を持つことを確認した．","subitem_description_type":"Other"}]},"item_4_description_8":{"attribute_name":"論文抄録(英)","attribute_value_mlt":[{"subitem_description":"This article describes a novel method for automatic speech recognition (ASR) based on kernel-based nonlinear classification. The new approach is obtained by substituting kernel function into inner-product performed in the dual problem of a learning problems of CD-HMMs formulated using minimum relative entropy discrimination (MRED). Unlike earlier work on sequential pattern recognition using kernel methods, our method can accurately model the three kinds of hierarchical dynamic patterns in CD-HMMs. Sequences of discrete-valued labels (string-level) or hidden states (state-level), as well as sequences of continuousvalued speech-derived frames (frame-level) can all be represented within the same kernel-based framework. Therefore, many efficient sequential pattern recognition algorithms for CD-HMMs (e.g. dynamic programming, Viterbi decoding, and the forward-backward algorithm) can be integrated into the new approach. This is an essential requirement for state-of-the-art speech recognition systems. We also describe a practical and effective optimization procedure for the proposed model, based on a cutting plane algorithm. The performance of the proposed method was evaluated in isolated phoneme recognition experiments. The method was found to be comparable with well-tuned conventional models.","subitem_description_type":"Other"}]},"item_4_biblio_info_10":{"attribute_name":"書誌情報","attribute_value_mlt":[{"bibliographicPageEnd":"6","bibliographic_titles":[{"bibliographic_title":"研究報告音声言語情報処理（SLP）"}],"bibliographicPageStart":"1","bibliographicIssueDates":{"bibliographicIssueDate":"2009-07-10","bibliographicIssueDateType":"Issued"},"bibliographicIssueNumber":"6","bibliographicVolumeNumber":"2009-SLP-77"}]},"relation_version_is_last":true,"weko_creator_id":"10"},"created":"2025-01-18T23:24:34.778782+00:00","updated":"2025-01-22T02:30:43.022928+00:00","id":62667}