@techreport{oai:ipsj.ixsq.nii.ac.jp:00232544,
 author = {井本, 桂右 and Keisuke, Imoto},
 issue = {74},
 month = {Feb},
 note = {機械学習技術の発展や計算機性能の向上により，音の分析・合成技術はこれまでにない速度で進化を遂げている．最新の技術では，静音環境下での音声認識性能はすでに人間を超えたと評されており，合成音声は人と区別できない品質に達している．そのような中で，環境音の分析・合成が新たな研究対象として扱われるようになってきた．従来，音声や楽音の分析・合成では，音の発生機構や時間周波数領域でのスパース性を活用した手法が検討されてきた．しかしながら，音声や楽音に限らないあらゆる音を対象とする環境音分析・合成技術では，音声や楽音の特徴を活用した従来手法をそのまま適用することが難しい．本稿では，環境音の分析・合成技術の基礎を紹介するとともに，環境音分析・合成ならではの課題や，環境音のドメイン知識を活用した分析・合成手法について議論する．また，当該分野の中心的役割を果たしている研究コミュニティ，Detection and Classification of Acoustic Scenes and Events (DCASE) の活動内容についても紹介し，最新の研究動向と今後の環境音の分析・合成技術の向かうべき方向性についても議論する．, With the development of machine learning techniques and improvements in computing power, audio analysis and synthesis technologies are advancing at an unprecedented rate. The latest technologies in speech processing have already achieved human speech recognition performance in quiet environments, and synthesized speech by the latest technologies has reached a quality indistinguishable from human speech. Under such circumstances, the analysis and synthesis of environmental sounds have become new research topics in audio and speech processing. In the analysis and synthesis of speech and musical sounds, many methods that take advantage of the sound generation mechanism and sparesness in the time-frequency domain have long been considered. However, it is difficult to utilize such domain knowledge for environmental sound analysis and synthesis. This paper introduces the fundamentals of environmental sound analysis and synthesis, and discusses research topics unique to environmental sound analysis and synthesis, as well as analysis and synthesis methods that utilize domain knowledge of environmental sounds. Then, the activities of Detection and Classification of Acoustic Scenes and Events (DCASE), a research community that plays a central role in this field, are also introduced, and the latest research trends and future directions of environmental sound analysis and synthesis are discussed.},
 title = {これから始める環境音分析・合成},
 year = {2024}
}