On 7th September 2022, Tokiya Ohmura (B4, at the time of presentation), Department of Information and Communication Engineering, received Student Presentation Award of the ASJ meeting.
About awarded research
A study of automatic evaluation of learners’ speech with robustness to background noise using a throat microphone.
Tokiya Ohmura, Yusuke Shozui, Kiichi Itoh, Daisuke Saito, Nobuaki Minematsu, Noriko Nakanishi
Your impression & future plan
Speech assessment technologies generally require clean speech, which is often difficult to obtain in a classroom. With a throat microphone, which detects not air particle vibrations but vibrations of the skin of the throat, only the target speaker’s voices can be recorded even with many surrounding learners. However, speech signals detected with a throat microphone are inevitably muffled. In this study, they are converted back to their clean and clear voices with DNN-based voice conversion techniques. As future work, further improvements of conversion performance will be needed.