{"links":{},"id":2003931,"metadata":{"_oai":{"id":"oai:ipsj.ixsq.nii.ac.jp:02003931","sets":["1164:4061:11910:1756715493468"]},"path":["1756715493468"],"owner":"80578","recid":"2003931","title":["台本読みの技術向上のためのスピーチ練習音声の自然発話・台本読み発話判定手法"],"pubdate":{"attribute_name":"PubDate","attribute_value":"2025-09-10"},"_buckets":{"deposit":"9c820026-ff07-48b4-a05c-379fbac86113"},"_deposit":{"id":"2003931","pid":{"type":"depid","value":"2003931","revision_id":0},"owners":[80578],"status":"published","created_by":80578},"item_title":"台本読みの技術向上のためのスピーチ練習音声の自然発話・台本読み発話判定手法","author_link":[],"item_titles":{"attribute_name":"タイトル","attribute_value_mlt":[{"subitem_title":"台本読みの技術向上のためのスピーチ練習音声の自然発話・台本読み発話判定手法","subitem_title_language":"ja"}]},"item_keyword":{"attribute_name":"キーワード","attribute_value_mlt":[{"subitem_subject":"UBI","subitem_subject_scheme":"Other"}]},"item_type_id":"4","publish_date":"2025-09-10","item_4_text_3":{"attribute_name":"著者所属","attribute_value_mlt":[{"subitem_text_value":"神戸大学大学院工学研究科"},{"subitem_text_value":"公立はこだて未来大学システム情報科学部情報アーキテクチャ学科"},{"subitem_text_value":"神戸大学大学院工学研究科"},{"subitem_text_value":"神戸大学大学院工学研究科"}]},"item_4_text_4":{"attribute_name":"著者所属(英)","attribute_value_mlt":[{"subitem_text_value":"Graduate School of Engineering, Kobe University","subitem_text_language":"en"},{"subitem_text_value":"School of Systems Information Science, Department of Media Architecture, Future University Hakodate","subitem_text_language":"en"},{"subitem_text_value":"Graduate School of Engineering, Kobe University","subitem_text_language":"en"},{"subitem_text_value":"Graduate School of Engineering, Kobe 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/2003931/files/IPSJ-UBI25087032.pdf","label":"IPSJ-UBI25087032.pdf"},"date":[{"dateType":"Available","dateValue":"2027-09-10"}],"format":"application/pdf","billing":["billing_file"],"filename":"IPSJ-UBI25087032.pdf","filesize":[{"value":"2.0 MB"}],"mimetype":"application/pdf","priceinfo":[{"tax":["include_tax"],"price":"660","billingrole":"5"},{"tax":["include_tax"],"price":"330","billingrole":"6"},{"tax":["include_tax"],"price":"0","billingrole":"36"},{"tax":["include_tax"],"price":"0","billingrole":"44"}],"accessrole":"open_date","version_id":"ad1c9536-7834-494d-8d05-2494055eff82","displaytype":"detail","licensetype":"license_note","license_note":"Copyright (c) 2025 by the Information Processing Society of Japan"}]},"item_4_creator_5":{"attribute_name":"著者名","attribute_type":"creator","attribute_value_mlt":[{"creatorNames":[{"creatorName":"谷重,翔生"}]},{"creatorNames":[{"creatorName":"渡邉,拓貴"}]},{"creatorNames":[{"creatorName":"寺田,努"}]},{"creatorNames":[{"creatorName":"塚本,昌彦"}]}]},"item_4_source_id_9":{"attribute_name":"書誌レコードID","attribute_value_mlt":[{"subitem_source_identifier":"AA11838947","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":"2188-8698","subitem_source_identifier_type":"ISSN"}]},"item_4_description_7":{"attribute_name":"論文抄録","attribute_value_mlt":[{"subitem_description":"スピーチやプレゼンテーションでは，発表者が話す内容をまとめた原稿に沿って話を進めることがある．発表者に台本を読み上げる技術が十分に備わっていないと，聞き手は「ただ台本を音読しているだけ」だと感じ，発表が印象に残らない．そのため，発表者は台本読み上げであっても自然な会話のように発話する技術が必要である．台本読み上げの個人練習では，自分の話し方が「台本読み」だと感じられる状態であるかを話し手が認識することが難しいため，話し方に対する聞き手の感じ方を提示することで台本読みの改善を支援できると考える．そこで本研究では，入力された音声が台本を読み上げていない自然な話し方であるか，台本を読み上げた話し方であるかを判定しフィードバックする台本読み上げの個人練習支援手法を提案する．本論文では，自然な話し方の音声と台本を読み上げた音声から取得したスペクトログラム画像を学習データとした深層学習によって，入力音声が自然な話し方であるか台本の読み上げであるかの分類を行うモデルを作成した．訓練データとテストデータを変えながら分類精度の評価実験を実施したところ，正解率が最高で97.5%，最低で66.0%，平均で79.7%となり，高い精度が得られる場合もあるが使用するデータによって精度に大きく差が出た．これに対し，AIボイスチェンジャーを用いて発話音声の声質を統一することで，声質が分類に与える影響を低減し分類精度を改善できるか評価したところ，正解率が最高で80.0%，最低で73.8%，平均で77.5%となり，使用データによる精度のばらつきを低減できた．","subitem_description_type":"Other"}]},"item_4_biblio_info_10":{"attribute_name":"書誌情報","attribute_value_mlt":[{"bibliographicPageEnd":"8","bibliographic_titles":[{"bibliographic_title":"研究報告ユビキタスコンピューティングシステム（UBI）"}],"bibliographicPageStart":"1","bibliographicIssueDates":{"bibliographicIssueDate":"2025-09-10","bibliographicIssueDateType":"Issued"},"bibliographicIssueNumber":"32","bibliographicVolumeNumber":"2025-UBI-87"}]},"relation_version_is_last":true,"weko_creator_id":"80578"},"created":"2025-09-01T10:23:18.142100+00:00","updated":"2025-09-09T01:52:49.263360+00:00"}