A study on custom-made augmented speech communication system based on higher-order statistics pursuit

2015 Fiscal Year Final Research Report

• Project/Area Number: 23240023
• Research Category: Grant-in-Aid for Scientific Research (A)
• Allocation Type: Single-year Grants
• Section: 一般
• Research Field: Perception information processing/Intelligent robotics
• Research Institution: The University of Tokyo (2014-2015); Nara Institute of Science and Technology (2011-2013)
• Principal Investigator: Saruwatari Hiroshi 東京大学, 情報理工学(系)研究科, 教授 (30324974)
• Co-Investigator(Kenkyū-buntansha): SHIKANO Kiyohiro 奈良先端科学技術大学院大学, 情報科学研究科, 名誉教授 (00263426) ; TODA Tomoki 奈良先端科学技術大学院大学, 情報科学研究科, 准教授 (90403328) ; KAWANAMI Hiromichi 奈良先端科学技術大学院大学, 情報科学研究科, 助教 (80335489) ; ONO Nobutaka 国立情報学研究所, 情報学プリンシプル研究系, 准教授 (80334259) ; MIYABE Shigeki 筑波大学, 大学院システム情報工学研究科, 助教 (50598745) ; MAKINO Shoji 筑波大学, 大学院システム情報工学研究科, 教授 (60396190) ; KOYAMA Shoichi 東京大学, 大学院情報理工学系研究科, 助教 (80734459)
• Project Period (FY): 2011-04-01 – 2016-03-31
• Keywords: 音声情報処理 / 統計的学習理論 / 音響信号処理

Outline of Final Research Achievements

In this study, we address an unsupervised custom-made augmented speech communication system based on the higher-order statistics pursuit. This system consists of two parts, namely, a binaural hearing aid using blind source separation and a speaking aid via speech conversion. The following results are obtained. (1) As the binaural hearing-aid system, we propose new algorithms for an accurate and fast blind source separation and statistical speech conversion, yielding a high quality speech enhancement system utilizing a fixed point of auditory perception. (2) As the speaking-aid system, a new robust speech conversion algorithm against a mismatch between speech database is proposed. The evaluation using real-world sound database shows the efficacy of the proposed method.

Free Research Field

知能情報処理