Toward construction of a mood induction technology according to cognitive feedback

2019 Fiscal Year Final Research Report

• Project/Area Number: 17K20011
• Research Category: Grant-in-Aid for Challenging Research (Exploratory)
• Allocation Type: Multi-year Fund
• Research Field: Human informatics and related fields
• Research Institution: Saga University
• Principal Investigator: Oshima Chika 佐賀大学, 理工学部, 客員研究員 (10395147)
• Co-Investigator(Kenkyū-buntansha): 中山 功一 佐賀大学, 理工学部, 准教授 (50418498)
• Project Period (FY): 2017-06-30 – 2020-03-31
• Keywords: 音声変換 / 音高と音量 / 機械学習 / 分類器

Outline of Final Research Achievements

The purpose of the research was to construct the technology that induces the speaker to misunderstand the mental state by arranging the speaker's voice and letting the speaker to listen to it in real time.
In this study, we compared our system, EPROCs (Emotional PROsody Conversion system) that converts volume and pitch in real time with a system that guides to any emotion by applying a filter that converts frequency to voice. We conducted an emotion classification experiment according to machine learning. As a result, a classifier that features sound quality data by applying the latter system could estimate the emotion with a higher accuracy rate than another classifier that features the pitch and volume data by applying the former system.

Free Research Field

ヒューマンコンピュータインタラクション

Academic Significance and Societal Importance of the Research Achievements

発話音声の音高と音量が変換された自分の声を聞く話者の気分が誘導されることを実験で明らかにしたいと考えた．しかし，話者の生の音声と変換した音声が二重に聞こえることと，変換した音声のわずかな遅延により，音声変換のみを気分誘導の要因とすることができなかった．そこで本研究では既存の周波数を変換するフィルタをかける手法と，本研究の音高と音量を変換する手法を比較する実験を行った．音高と音量を変換する手法は感情分類の正解率が低くなることがわかった．