InterSpeech 2021

Vocal-tract models to visualize the airstream of human breath and droplets while producing speech
(3 minutes introduction)

Takayuki Arai (Sophia University, Japan)
Due to the COVID-19 pandemic, visualizing the airstream of human breath during speech production has become extremely important from the viewpoint of preventing infection. In addition, visualizing droplets and the larger drops expelled when we speak consonantal sounds may help for the same reason. One visualization technique is to pass a laser sheet through the droplet cloud produced by a human speaker. However, the laser poses certain health risks for human beings. Therefore, we developed an alternative method to passing a laser against a human body in which we utilize physical models of the human vocal tract. First, we tested a head-shaped model with a lung model from our previous study to visualize the exhaled breath during vowel production (with and without a mask). Then, we implemented an extended version of the anatomical-type vocal-tract model introduced in our previous study. With this newly developed model, lips are made of the same flexible material that was used to form the tongue part in the previous model. We also attached these lips to another previous model for producing sounds including /b/. Finally, the lip models were tested to visualize the droplet cloud including expelled drops present while producing a bilabial plosive sound.