{"links":{},"id":197814,"metadata":{"_oai":{"id":"oai:ipsj.ixsq.nii.ac.jp:00197814","sets":["1164:5064:9706:9832"]},"path":["9832"],"owner":"44499","recid":"197814","title":["深層音声生成モデルと同時対角化可能な空間相関行列に基づく高速マルチチャネル音声強調"],"pubdate":{"attribute_name":"公開日","attribute_value":"2019-06-15"},"_buckets":{"deposit":"a9996d42-4f90-4ccc-815a-563c86be633b"},"_deposit":{"id":"197814","pid":{"type":"depid","value":"197814","revision_id":0},"owners":[44499],"status":"published","created_by":44499},"item_title":"深層音声生成モデルと同時対角化可能な空間相関行列に基づく高速マルチチャネル音声強調","author_link":["475365","475366","475367","475364"],"item_titles":{"attribute_name":"タイトル","attribute_value_mlt":[{"subitem_title":"深層音声生成モデルと同時対角化可能な空間相関行列に基づく高速マルチチャネル音声強調"}]},"item_keyword":{"attribute_name":"キーワード","attribute_value_mlt":[{"subitem_subject":"ポスターセッション1","subitem_subject_scheme":"Other"}]},"item_type_id":"4","publish_date":"2019-06-15","item_4_text_3":{"attribute_name":"著者所属","attribute_value_mlt":[{"subitem_text_value":"理化学研究所AIP／京都大学大学院情報学研究科"},{"subitem_text_value":"理化学研究所AIP"},{"subitem_text_value":"産業技術総合研究所"},{"subitem_text_value":"理化学研究所AIP／京都大学大学院情報学研究科"}]},"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/197814/files/IPSJ-MUS19123029.pdf","label":"IPSJ-MUS19123029.pdf"},"date":[{"dateType":"Available","dateValue":"2021-06-15"}],"format":"application/pdf","billing":["billing_file"],"filename":"IPSJ-MUS19123029.pdf","filesize":[{"value":"990.2 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":"21"},{"tax":["include_tax"],"price":"0","billingrole":"44"}],"accessrole":"open_date","version_id":"434342fc-6704-405b-885e-0d75c32e7124","displaytype":"detail","licensetype":"license_note","license_note":"Copyright (c) 2019 by the Information Processing Society of Japan"}]},"item_4_creator_5":{"attribute_name":"著者名","attribute_type":"creator","attribute_value_mlt":[{"creatorNames":[{"creatorName":"關口, 航平"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"Aditya, Arie Nugraha"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"坂東, 宜昭"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"吉井, 和佳"}],"nameIdentifiers":[{}]}]},"item_4_source_id_9":{"attribute_name":"書誌レコードID","attribute_value_mlt":[{"subitem_source_identifier":"AN10438388","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_source_id_11":{"attribute_name":"ISSN","attribute_value_mlt":[{"subitem_source_identifier":"2188-8752","subitem_source_identifier_type":"ISSN"}]},"item_4_description_7":{"attribute_name":"論文抄録","attribute_value_mlt":[{"subitem_description":"本稿では，フルランク空間相関行列に基づくマルチチャネル音源分離を高速に実行するための，収束保証付きの汎用的なアルゴリズムについて述べる．代表的な音源分離法であるマルチチャネル非負値行列因子分解 (MNMF) では，各音源スペクトログラムのパワースペクトル密度が低ランク構造を持つと仮定している．音声スペクトログラムに対してこの仮定は成り立たないため，最近では，音声に対しては事前学習した深層生成モデルを用い，雑音に対してはNMFに基づく低ランクモデルを用いた音声強調法が提案されている．これらの手法は，フルランクの空間相関行列を直接取り扱う上で計算量が大きく，実用上の課題となっていた．本稿では，各周波数において，各音源に対応する空間相関行列が同時対角化可能であるという制約のもとでは，観測スペクトログラムを線形変換することで，各チャネルを独立化でき，共分散行列演算が回避できることを示す．具体的には，独立ベクトル分析 (IVA) で提案された反復射影法 (IP) を用いた変換行列の推定と，変換後の空間での非負値テンソル分解 (NTF) との反復を行うことで，収束保証付きの最適化アルゴリズムを導出できる．提案する同時対角化可能フルランク空間モデルは，独立低ランク行列分析 (ILRMA) で用いられるランク 1 空間モデルと深い関係がある．実験では，ILRMA と同等の計算量に削減しつつ，初期値依存性が小さく，より高精度な音声強調ができることを確認した．","subitem_description_type":"Other"}]},"item_4_biblio_info_10":{"attribute_name":"書誌情報","attribute_value_mlt":[{"bibliographicPageEnd":"6","bibliographic_titles":[{"bibliographic_title":"研究報告音楽情報科学（MUS）"}],"bibliographicPageStart":"1","bibliographicIssueDates":{"bibliographicIssueDate":"2019-06-15","bibliographicIssueDateType":"Issued"},"bibliographicIssueNumber":"29","bibliographicVolumeNumber":"2019-MUS-123"}]},"relation_version_is_last":true,"weko_creator_id":"44499"},"created":"2025-01-19T01:02:14.022607+00:00","updated":"2025-01-19T22:11:54.933516+00:00"}