{"created":"2025-10-20T05:28:20.750007+00:00","metadata":{"_oai":{"id":"oai:ipsj.ixsq.nii.ac.jp:02004841","sets":["934:6391:11874:1757990589928"]},"path":["1757990589928"],"owner":"80578","recid":"2004841","title":["ウェアラブルセンサのスペクトル可視化による転倒検出手法の検討"],"pubdate":{"attribute_name":"PubDate","attribute_value":"2025-10-24"},"_buckets":{"deposit":"a0c4b12e-fe9c-4c8d-9b76-e2ca05d4b1f6"},"_deposit":{"id":"2004841","pid":{"type":"depid","value":"2004841","revision_id":0},"owners":[80578],"status":"published","created_by":80578},"item_title":"ウェアラブルセンサのスペクトル可視化による転倒検出手法の検討","author_link":[],"item_titles":{"attribute_name":"タイトル","attribute_value_mlt":[{"subitem_title":"ウェアラブルセンサのスペクトル可視化による転倒検出手法の検討","subitem_title_language":"ja"},{"subitem_title":"Investigation of Fall Detection Method by Spectral Visualization of Wearable Sensors","subitem_title_language":"en"}]},"item_keyword":{"attribute_name":"キーワード","attribute_value_mlt":[{"subitem_subject":"[コンシューマ・システム論文] 転倒検出，異常検知，センサデータ，スカログラム，畳み込みオートエンコーダ，OC-SVM","subitem_subject_scheme":"Other"}]},"item_type_id":"3","publish_date":"2025-10-24","item_3_text_3":{"attribute_name":"著者所属","attribute_value_mlt":[{"subitem_text_value":"静岡大学大学院総合科学技術研究科情報学専攻／ローランドディー.ジー.株式会社"},{"subitem_text_value":"静岡大学学術院情報学領域／静岡大学グリーン科学技術研究所"}]},"item_3_text_4":{"attribute_name":"著者所属(英)","attribute_value_mlt":[{"subitem_text_value":"Department of Informatics, Graduate School of Integrated Science and Technology, Shizuoka University / Roland DG Corporation","subitem_text_language":"en"},{"subitem_text_value":"College of Informatics, Academic Institute, Shizuoka University / Research Institute of Green Science and Technology, Shizuoka 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/2004841/files/IPSJ-TCDS1503004.pdf","label":"IPSJ-TCDS1503004.pdf"},"date":[{"dateType":"Available","dateValue":"2027-10-24"}],"format":"application/pdf","billing":["billing_file"],"filename":"IPSJ-TCDS1503004.pdf","filesize":[{"value":"2.8 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":"47"},{"tax":["include_tax"],"price":"0","billingrole":"44"}],"accessrole":"open_date","version_id":"68c586c8-4704-42a1-adfb-4aaf2b9890d7","displaytype":"detail","licensetype":"license_note","license_note":"Copyright (c) 2025 by the Information Processing Society of Japan"}]},"item_3_creator_5":{"attribute_name":"著者名","attribute_type":"creator","attribute_value_mlt":[{"creatorNames":[{"creatorName":"束野,通洋"}]},{"creatorNames":[{"creatorName":"峰野,博史"}]}]},"item_3_creator_6":{"attribute_name":"著者名(英)","attribute_type":"creator","attribute_value_mlt":[{"creatorNames":[{"creatorName":"Michihiro Tsukano","creatorNameLang":"en"}]},{"creatorNames":[{"creatorName":"Hiroshi Mineno","creatorNameLang":"en"}]}]},"item_3_source_id_9":{"attribute_name":"書誌レコードID","attribute_value_mlt":[{"subitem_source_identifier":"AA12628043","subitem_source_identifier_type":"NCID"}]},"item_resource_type":{"attribute_name":"資源タイプ","attribute_value_mlt":[{"resourceuri":"http://purl.org/coar/resource_type/c_6501","resourcetype":"journal article"}]},"item_3_source_id_11":{"attribute_name":"ISSN","attribute_value_mlt":[{"subitem_source_identifier":"2186-5728","subitem_source_identifier_type":"ISSN"}]},"item_3_description_7":{"attribute_name":"論文抄録","attribute_value_mlt":[{"subitem_description":"高齢者の転倒は死に至る重大な要因であり，転倒後の迅速な対応のために，深層学習を活用した転倒検出技術の研究が進展している．ここで，検出精度の向上には大量の学習データが必要となるが，実際の転倒データを収集するには被験者の危険をともない，データ収集が困難である．本研究では，異常検知手法に着目し，比較的容易に収集可能な正常な行動データのみを用いた転倒検出手法を提案する．具体的には，加速度センサとジャイロセンサの時系列データをスペクトル可視化し，畳み込みオートエンコーダを用いて特徴量を抽出した後，1クラスサポートベクターマシンを用いて，正常行動と転倒の二クラス分類を行った．ここで，複数のセンサに対応するため，センサごとに畳み込みオートエンコーダを構築した．評価実験では，元のセンサデータ画像，スペクトログラム，スカログラムを用い，センサごとに構築した単体モデルおよび複数モデルの性能を比較した．その結果，スカログラムを用いた複数モデル手法によって，F1-scoreが0.915となる高精度を確認した．さらに，Azureを用いたプロトタイプシステムを開発し，実効性も検証した．複雑な行動に対しては課題が残されるが，単純な行動には有効であることが明らかとなった．","subitem_description_type":"Other"}]},"item_3_description_8":{"attribute_name":"論文抄録(英)","attribute_value_mlt":[{"subitem_description":"Falls among the elderly are a serious cause of death, and research on fall detection using deep learning has been progressing in order to respond quickly after a fall. However, collecting actual fall data is difficult due to the risk to the subjects. In this study, we focus on anomaly detection methods and propose a fall detection method using only normal behavioral data, which can be collected relatively easily. Specifically, the time-series data from the accelerometer and gyro sensor were spectrally visualized, features were extracted using a convolutional autoencoder, and a one-class support vector machine was used to perform two-class classification: normal behavior and falls. In order to support multiple sensors, a convolutional autoencoder was constructed for each sensor. In the evaluation experiment, we compared the performance of the single and multiple models built for each sensor using the original sensor data images, spectrograms, and scalograms. As a result, we confirmed that the multiple model method using the scalograms achieved a high accuracy with an F1-score of 0.915. Furthermore, we developed a prototype system using Azure and verified its effectiveness. The system was found to be effective for simple behaviors, although there are still some issues to be solved for complex behaviors.","subitem_description_type":"Other"}]},"item_3_biblio_info_10":{"attribute_name":"書誌情報","attribute_value_mlt":[{"bibliographicPageEnd":"28","bibliographic_titles":[{"bibliographic_title":"情報処理学会論文誌コンシューマ・デバイス＆システム（CDS）"}],"bibliographicPageStart":"14","bibliographicIssueDates":{"bibliographicIssueDate":"2025-10-24","bibliographicIssueDateType":"Issued"},"bibliographicIssueNumber":"3","bibliographicVolumeNumber":"15"}]},"relation_version_is_last":true,"weko_creator_id":"80578"},"id":2004841,"updated":"2025-10-23T06:40:32.347576+00:00","links":{}}