Study of sound quality improvement for body-conducted speech with acoustic clustering

2019 Fiscal Year Final Research Report

• Project/Area Number: 15K00245
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Research Field: Perceptual information processing
• Research Institution: Hiroshima City University
• Principal Investigator: Nakayama Masashi 広島市立大学, 情報科学研究科, 講師 (20508163)
• Co-Investigator(Kenkyū-buntansha): 籠宮 隆之 国立研究開発法人産業技術総合研究所, その他部局等, 研究員 (10528269) ; 石光 俊介 広島市立大学, 情報科学研究科, 教授 (70300621) ; 中川 誠司 国立研究開発法人産業技術総合研究所, その他部局等, 研究員 (70357614)
• Project Period (FY): 2015-04-01 – 2020-03-31
• Keywords: 骨伝導音 / 体内伝導音 / 音声 / 皮膚伝搬 / 三次元変位計測

Outline of Final Research Achievements

The acoustic similarity was obtained from a database containing body-conducted speech and airborne speech, and the research goal was to achieve improvement for sound quality of body-conducted speech by acoustic clustering focusing on vocabulary and articulation methods. However, it was confirmed that contact conditions between pickup and skin influence criticality of sound rather than spoken vocabulary and articulation method. Therefore, we focused on three-dimensional displacement measurement which has less affected by the contact condition between the pickup and skin, and confirmed the relationship between speech and body-conducted speech from skin conduction. It can be said that the possibility of acoustic clustering was confirmed from the bias of skin vibration when Japanese vowels were uttered, because skin-conduction could be measured by three-dimensional displacement measurement.

Free Research Field

音声信号処理

Academic Significance and Societal Importance of the Research Achievements

スマートフォンにおける音声検索や講義要約など音声インタフェースによる様なタスクやドキュメント処理が広く普及してきた．しかしながら，音声は気導雑音の影響を受けやすいことから雑音に頑健な音声入力が求められている．骨伝導音は皮膚伝搬であるため気導雑音の影響を受けにくい性質があるため，雑音環境下での音声採取法として有用である．一方で，2kHz以上における周波数成分の減衰が著しいため，骨伝導音の伝搬解明や音質改善が求められている．本研究では骨伝導音の伝搬解明と音質改善に資する基礎研究として位置付けられ，その学術的また社会的意義が大きいものであるといえる．