A "easy hearing" public address system with microphone network and high quality speech synthesis

2021 Fiscal Year Final Research Report

• Project/Area Number: 19K15146
• Research Category: Grant-in-Aid for Early-Career Scientists
• Allocation Type: Multi-year Fund
• Review Section: Basic Section 23020:Architectural environment and building equipment-related
• Research Institution: Muroran Institute of Technology
• Principal Investigator: Kobayashi Yosuke 室蘭工業大学, 大学院工学研究科, 助教 (10735103)
• Project Period (FY): 2019-04-01 – 2022-03-31
• Keywords: 屋外拡声システム / 音声了解度 / 音声品質 / 音声合成

Outline of Final Research Achievements

In order to improve the listening of outdoor public address (PA) system, we conducted a basic study on a speech quality prediction using multiple microphones and a speech synthesis that feeds back the PA speech quality. First, we developed a prototype using a microphone array for PA speech measurement. Next, we developed a speech quality prediction model that uses deep learning. Finally, for speech synthesis, we implemented a statistical speech synthesis model that complements the high frequencies filtered by wireless communication. In this study, only these unit tests were performed. However, if a PA system that cooperates in a complex manner is realized, it is thought that better hearing will be realized.

Free Research Field

音響工学

Academic Significance and Societal Importance of the Research Achievements

屋外拡声に聴き取りにくいことがあることは多くの人が理解していることである。なぜ聴こえにくいかの要因は騒音や風の影響，残響（響き）の影響など多数ある。この問題を解決するには，どの程度聞き取りにくいのかを定量化する事と，この値をフィードバックしながら聴こえやすく音声を処理し直すことの二つが必要である。本研究では，拡声音声の計測（録音），計測信号の品質評価，品質値の音声合成へのフィードバックを一体的に行うシステムの基礎検討をおこなった。