Human Learning from Ears: Research on Otopecia

2016 Fiscal Year Final Research Report

• Project/Area Number: 26540092
• Research Category: Grant-in-Aid for Challenging Exploratory Research
• Allocation Type: Multi-year Fund
• Research Field: Perceptual information processing
• Research Institution: National Institute of Informatics
• Principal Investigator: Satoh Ken 国立情報学研究所, 情報学プリンシプル研究系, 教授 (00271635)
• Co-Investigator(Renkei-kenkyūsha): AIZAWA Akiko 国立情報学研究所, コンテンツ科学研究系, 教授 (90222447) ; MIYAO Yusuke 国立情報学研究所, コンテンツ科学研究系, 准教授 (00343096) ; BONO Mayumi 国立情報学研究所, コンテンツ科学研究系, 准教授 (50418521)
• Project Period (FY): 2014-04-01 – 2017-03-31
• Keywords: 音声合成 / 話者適応 / ディープラーニング / Otopedia

Outline of Final Research Achievements

For human learning from ears, we developed a method of making speaker-adaptive voice synthesis system. We showed that we can use deep-learning for the purpose with small amount of voice data as the previous Hidden Markov models. We also made a iOS application of audio book reader using the above method and demonstrated that the application could read digital books in a ePub format using speaker's voice. We could use this system for analysis of engagement time and learning performance using various voices.

Free Research Field

人工知能基礎