Скачать или смотреть SottoVoce: An Ultrasound Imaging-Based Silent Speech Interaction Using Deep Neural Networks

Abstract
The availability of digital devices operated by voice is rapidly expanding. However, the usage situation of voice interfaces is still restricted. For example, speaking in public places becomes an annoyance to surrounding people, and secret information should not be uttered. Environmental noise may reduce the accuracy of the speech recognition. To address these limitations, SottoVoce detects a user’s unuttered voice. From internal information observed by an ultrasonic imaging sensor attached to the underside of the jaw, our proposed system recognizes the utterance contents without the user actually uttering voice. Our proposed deep neural network model is used to obtain acoustic features from a sequence of ultrasound images. We confirmed that audio signals generated by our system can control existing smart speakers. We also observed that a user can adjust their oral movement to learn and improve accuracy of their voice recognition.

・Naoki Kimura, Michinari Kono, Jun Rekimoto. SottoVoce: An Ultrasound Imaging-Based Silent Speech Interaction Using Deep Neural Networks, ACM CHI 2019

・暦本 純一（東大／ソニーCSL），木村 直紀，河野 通就（東大）, SottoVoce: 超音波画像と深層学習による無発声音声インタラクション, インタラクション2019

Project Page
https://lab.rekimoto.org/projects/sot...