Estimating confidence in voices using crowdsourcing for alleviating tension with altered auditory feedback
Naruse Kana, Shinnosuke Takamichi, Tomohiro Tanikawa, Shigeo Yoshida, Takuji Narumi, Michitaka Hirose · 2019
People tend to face difficulties while speaking, owing to excessive tension. To relieve this tension, we propose a method to alter their voice such that it appears more confident and feed this voice back to the speaker in real time. As determining the appropriate parameters for voice processing is difficult, we gathered data on the perception of confidence in voices through crowdsourcing and constructed a model for estimating confidence scores from voice-processing parameters. An analysis of the model showed that although the coefficient of determination was not considerably large, the inflection of speech tended to affect the perception of confidence.