Sound event localization and detection utilizing overlapping end-to-end learning

• Item type: Symposium(1)
• 公開日: 2023-12-20
• タイトル: Sound event localization and detection utilizing overlapping end-to-end learning
• タイトル / 言語: en
• タイトル: Sound event localization and detection utilizing overlapping end-to-end learning
• 言語: eng
• 資源タイプ / 資源タイプ識別子: http://purl.org/coar/resource_type/c_5794
• 資源タイプ: conference paper
• 著者所属: Tokyo Institute of Technology
• 著者所属: Keio University
• 著者所属: Honda Research Institute Japan Co., Ltd.
• 著者所属: Tokyo Institute of Technology／Honda Research Institute Japan Co., Ltd.
• 著者所属: Tokyo Institute of Technology
• 著者所属: Tokyo Institute of Technology
• 著者所属: Keio University
• 著者所属(英): en
• 著者所属(英): Tokyo Institute of Technology
• 著者所属(英): en
• 著者所属(英): Keio University
• 著者所属(英): en
• 著者所属(英): Honda Research Institute Japan Co., Ltd.
• 著者所属(英): en
• 著者所属(英): Tokyo Institute of Technology / Honda Research Institute Japan Co., Ltd.
• 著者所属(英): en
• 著者所属(英): Tokyo Institute of Technology
• 著者所属(英): en
• 著者所属(英): Tokyo Institute of Technology
• 著者所属(英): en
• 著者所属(英): Keio University
• 著者名: Yanke, Long; Riku, Yasuda; Yui, Sudo; Katsutoshi, Itoyama; Kazuhiro, Nakadai; Kenji, Nishida; Hideharu, Amano
• 著者名(英): Yanke, Long; Riku, Yasuda; Yui, Sudo; Katsutoshi, Itoyama; Kazuhiro, Nakadai; Kenji, Nishida; Hideharu, Amano
• 論文抄録 / 内容記述タイプ: Other
• 論文抄録 / 内容記述: This paper presents efficient end-to-end deep learning for sound event localization and detection by sharing a part of the models called overlapping end-to-end learning, which can be trained with a small amount of data compared to normal end-to-end learning. We demonstrate its superior accuracy compared to traditional cascade integration, achieving a 3.3-point increase in classifying each of the mixed sound sources.
• 論文抄録(英) / 内容記述タイプ: Other
• 論文抄録(英) / 内容記述: This paper presents efficient end-to-end deep learning for sound event localization and detection by sharing a part of the models called overlapping end-to-end learning, which can be trained with a small amount of data compared to normal end-to-end learning. We demonstrate its superior accuracy compared to traditional cascade integration, achieving a 3.3-point increase in classifying each of the mixed sound sources.
• 書誌情報: Proceedings of Asia Pacific Conference on Robot IoT System Development and Platform; 巻 2023, p. 63-64, 発行日 2023-12-20
• 出版者 / 言語: ja
• 出版者: 情報処理学会